Remon Romp - Thesis (Final) HQ

Delft University of Technology
Master Thesis
Installation-effects of suction caissons

in non-standard soil conditions
- Final report -
Remon H. Romp
1373722

Faculty of Civil Engineering
and Geosciences
Section Geo-Engineering
8th October 2013

ii
Administrative Data
Student
Remon Romp
remon romp@hotmail.com
+316 123 88 376
Committee
Prof. Ir. A.F. van Tol - Chairman
Faculty of Civil Engineering and Geosciences, Room KG 00.140
A.F.vanTol@tudelft.nl
+3115 278 8592
Dr. Ir. J. Dijkstra - Daily supervisor University

Faculty of Civil Engineering and Geosciences, Room KG 00.510
jelke.dijkstra@chalmers.se
+463 177 22 120
Ir. S. Frankenmolen - Daily supervisor Company

Geotechnical Engineer
Dr. Ir. K.J. Bakker - External member

Faculty of Civil Engineering and Geosciences, Room HG 3.77.1
K.J.Bakker@tudelft.nl
+3115 278 5075
Addresses
Delft University of Technology SBM Offshore

Faculty of Civil Engineering and Geosciences Engineering Mooring Systems
Departement Geo-Technology Geotechnical Departement
Stevinweg 1 24 Avenue de Fontvieille

2628 CN Delft P.O. Box 199
The Netherlands MC 98007 Monaco Cedex
www.ge.citg.tudelft.nl www.sbmoffshore.com
iii
iv ADMINISTRATIVE DATA
Preface
This report is the final exercise of the Master program Geotechnical Engineering at the Delft Univer-
sity of Technology. The investigated subject ”Installation effects of suction caissons in non-standard
soil conditions” originates from the company SBM Offshore (SBM), a leading offshore engineering
company in floating production and mooring systems. During the graduation progress I was located
at the Geotechnical department of SBM Offshore in Monaco for a period of 3 months to obtain in-
formation from various practical experiences.
The report focusses on the installation-aspects of suction caissons in layered soil conditions, es-
pecially where sand is overlaid by clay. After literature review and experiments on plug uplift and
cracking, an assessment of the plug uplift is presented.
Throughout the entire research project I have received lots of advice, support and help from many
people. That is why I would like to seize the opportunity to express my sincere thanks to everyone
who has contributed to the completion of this project.
In the first place I would like to express my gratitude to the members of my committee, who
supported me throughout this entire research project. To my supervisor Sebastiaan Frankenmolen,
for his patience, his enthusiasm and guidance. Numerous discussions helped me to get a good un-
derstanding of all relevant geotechnical phenomena. To Jelke Dijkstra, my supervisor at the Delft
University of Technology, for his interest, constructive criticism and the effort he took to teach me
the tricks of physical modeling. I have appreciated the feedback I got, as even in weekends my ques-
tions were answered. To Professor Frits van Tol and Klaas-Jan Bakker, for answering my questions
whenever needed and have guided me in the right direction.
Besides my committee, I would like to thank my colleagues at SBM Offshore, for their interest
in my research project and their varying contributions to the completion of my thesis. Furthermore
my thanks go to Han de Visser, for his practical and cheerful assistance during my model tests. Last
but not least, I would like to thank my family and friends, who had to listen for eight months to my,
undoubtedly very interesting, stories about suction caissons.
8th October 2013,
Remon Hendrik Romp
v
vi PREFACE
Abstract
Nowadays suction caissons are extensively used for anchoring large offshore floating facilities. Install-
ation of suction caissons in coarse grained soils (high permeability) is achieved by induced seepage
flow in the soil. This reduces the soil resistance at the caisson tip, which contributes to the installa-
tion process. Installation in fine grained materials with a low permeability (silt or clay), is achieved
by suction (relative to seabed pressure) within the caisson, which forces the caisson to embed. As
a result the installation in layered soils, especially sand overlaid by clay, is expected to be more
challenging since the reduction of tip resistance is restrained from flow restrictions in the clay layer.
It is unclear to what extent the upper low permeable layer will impact the seepage flow and thus the
desired tip reduction.
During the installation in layered soils one of the main potential obstacles is excessive plug uplift,
which could occur at certain applied suction pressures. The objective of the presented research is to
assess the governing mechanisms of plug stability during installation of a suction caisson in layered
soils where sand is overlaid by clay.
It was found from case studies on installation in layered soils that water-injection-devices (WID)
were used. Subsequently, two general types of failure mechanisms of plug stability for installation
of suction caissons in sand overlying by clay were identified; plug uplift and plug cracking. Further
analysis showed that the failure mechanism depends on the dimensions of both the caisson (diameter
D) and the clay plug (layer thickness z ). Furthermore, basic experiments executed as part of this
research showed that plug cracking is an unlikely failure mechanism for standard aspect ratios (i.e.
D/z < 6)
Assessment of the plug stability was done by introducing two extreme cases; a fixed plug and a
moving plug during installation. Stability of a fixed plug was assessed by introducing a fracture-
ratio, which determines the crack-requirement for installation without uplift. When plug uplift can
be accommodated to a certain extend (i.e. a moving plug), the relation between the plug uplift
velocity and the installation-rate should be assessed. Based on analytical analyses, it is concluded
that for low installation rates the total amount of uplift is higher compared to fast installation rates.
Finally, the contribution of reverse end bearing (REB) as part of the plug stability was investig-
ated for layered soils. The uplift mechanism at the interface of the plug (i.e. the clay-sand interface)
consists of the process of suction mobilization and suction release within the sand. It was found that
the magnitude and duration of the temporary resistance (i.e. the reverse end bearing) depends on
the dissipation of negative excess pore pressures in time. This shows that for a typical installation
period (1 - 6 hours) and hydraulic conductivities lower than 1e−5 m/s, temporary capacity due to
delayed inflow of water can be considered.
The analytical work was verified with the results of two analysed field cases and numerical ana-
lysis. In both field cases water injection devices were implemented as mitigating measure (i.e. to
decrease tip resistance) to ensure penetration of the suction caisson in dense sands. However, this
vii
viii ABSTRACT
system is at present still not a proven technology and back-calculations did not clearly indicate the
benefit of these devices. The contribution (or the absence) of REB was recognized for both practical
and numerical back-calculations. This indicates that the magnitude of REB is highly dependent on
time-effects (i.e. the permeability of the sand and installation rates).
Samenvatting
Zuigpalen worden tegenwoordig veelal gebruikt als ankerpunt voor diverse drijvende voorzieningen
in de offshore industrie. Installatie van een zuigpaal in zandige grond (met een hoge doorlatend-
heid) wordt gerealiseerd door de opgewekte lekstroom in de grond, wat tevens een reductie van de
paalpuntweerstand genereert. Voor installatie in fijnere grond met een lagere doorlatendheid (klei of
silt) wordt de relatieve onderdruk in het caisson gebruikt om een kracht te genereren die de paal de
grond in drukt. Installaties in gelaagde gronden, waar zand bedekt is met een kleilaag, is complexer
omdat de lekstroom verhinderd wordt door de slecht doorlatende kleilaag. Het is onduidelijk wat de
invloed van deze slecht doorlatende toplaag is op de lekstroom en tevens op de vereiste reductie van
de paalpuntweerstand.
Extreme opbarsting van de klei plug vormt een potentieel probleem bij hogere drukken gedurende
installatie van de paal. Dit rapport heeft ten doel om de belangrijkste mechanismen (m.b.t. de sta-
biliteit van de plug) te onderzoeken tijdens installatie van een zuigpaal in de gelaagde grond conditie,
waar zand bedekt is met een kleilaag.
Uit praktijkstudies is gebleken dat een water-injection-device (WID) wordt gebruikt om zuigpalen te
installeren in gelaagde gronden. Daarnaast zijn er twee faalmechnismes beschreven voor een grond-
profiel waar zand bedekt is met een kleilaag; opbarsten van de kleilaag & scheuring van de kleilaag.
Toegepaste analyses tonen aan dat het faalmechanisme afhankelijk is van de gekozen dimensies, zoals
paal diameter (D) en laagdikte van de klei (z). Daarnaast is door middel van eenvoudige experi-
menten aangetoond dat scheuring van de kleilaag optreedt voor dimensies met D/z > 6.
De stabiliteit van de kleilaag is beschouwd door twee condities te introduceren; een gefixeerde plug of
een bewegende plug. In het eerste geval dient de doorlatendheid van de kleilaag vergroot te worden
door scheuring van de laag, wat in een percentage is uitgedrukt als fracture-ratio. Voor een bewe-
gende plug geldt dat de relatie tussen opbarsten van de klei en de snelheid van penetratie van belang
zijn. Op basis van analytische vergelijkingen kan worden geconcludeerd dat voor lagere penetraties-
nelheden de verwachte opwaartse verplaatsing van de kleilaag groter is.
Ten slotte is de bijdrage van reverse end bearing (REB) als deel van de stabiliteit van de plug
onderzocht voor gelaagde gronden. Het opbarst mechanisme bij de interface van de kleilaag en het
zand bestaat uit het mobilizeren van zuiging, gevolgd door het wegvloeien ervan in het onderliggende
zand. De grootte en duur van deze tijdelijke weerstand tegen opbarsten hangt af van de dissipatie
van negatieve wateroverspanningen in het onderliggende zand. Voor een gebruikelijke installatieduur
(1 - 6 uur) en doorlatendheden lager dan 1e−5 m/s, kan een tijdelijke capaciteit door de vertraagde
grondwaterstroming worden beschouwd.
De analytische resultaten zijn op basis van de gestelde randvoorwaarden geverifieerd door middel
van twee praktijkvoorbeelden en numerieke analyses. Voor beide praktijkgevallen zijn WID’s als
middel gebruikt om de paal in de dichtgepakte onderliggende zandlaag te installeren. Desondanks
is de effectiviteit van deze WID’s is niet bewezen en ook met berekeningen niet aantoonbaar. De
ix
x SAMENVATTING
bijdrage (of afwezigheid) van REB kon worden aangetoond voor beide praktijkvoorbeelden en nu-
merieke berekeningen. Beide analyses tonen aan dat de grootte van REB sterk afhankelijk is van
tijdseffecten, zoals doorlatendheid van het onderliggende zand en penetratiesnelheid.
Contents
Administrative Data iii
Preface v
Abstract vii
Samenvatting ix
Symbols xxi
1 Introduction 1
1.1 Background information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Problem description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Purpose of research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Readers manual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Installation in layered soils 5

2.1 Installation of a suction caisson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Reduction of tip resistance in layered soils . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Seepage flow in sand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.3 Effects of penetration-rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Initial phase: Self-weight penetration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Effective stress-approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 CPT-approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Penetration in layered soils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.4 Comparison of approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Second phase: Suction-assisted penetration . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.1 Critical suction in layered soils . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.2 Reduction of soil resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.3 Comparison of methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3 Plug response during installation 21

3.1 Theory from literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.1 Theory of plug uplift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.2 Theory of plug cracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.1.3 Reference data of plug response . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.1.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
xi
xii Table of contents
3.2 Experiments on plug response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.2.1 Scaling laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.2 Test set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.4 Conclusions & discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4 Assessment of plug stability 37

4.1 Installation with a stable plug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.1.1 Seepage through clay plug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.1.2 Suction below the plug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.1.3 Introducing the fracture ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.1.4 Effect of layer thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.1.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2 Installation with a moving plug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.1 Principle of volume continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.2 Suction below the plug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.2.3 Uplift-rate of the plug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5 Modelling components of plug uplift 47

5.1 Mechanisms at the interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.2 Reverse end bearing in sand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2.1 Time dependency of reverse end bearing . . . . . . . . . . . . . . . . . . . . . . 50
5.2.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.3 Modelling the plug uplift mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.3.1 Uplift by suction only . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.3.2 Uplift by seepage inflow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3.3 Combination of suction and seepage inflow . . . . . . . . . . . . . . . . . . . . 55
5.3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6 Verification of the theory 59

6.1 Case A: Curlew FPSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.1.1 Measured suction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.1.2 Plug uplift back-calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.1.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.2 Case B: Gullfaks C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.2.1 Measured suction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.2.2 Plug uplift back-calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.2.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.3 Numerical calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.3.1 Calculation method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.3.2 Modelling plug uplift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.3.4 Discussion & conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
xiii
7 Conclusions & Recommendations 81

7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.2 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Appendices 89
A Implementation of effective stress and CPT-approaches 89

A.1 Calculations according to effective stress-approach . . . . . . . . . . . . . . . . . . . . 89
A.2 Calculations according to CPT-approach . . . . . . . . . . . . . . . . . . . . . . . . . . 90
A.3 Correlation of effective stress- and CPT-approaches . . . . . . . . . . . . . . . . . . . . 90
B Determination of end-bearing coefficient 93
C Prediction method to determine cracking of the plug 95
D Installation of a suction caisson 97

D.1 Steady-state suction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
D.2 Elastic plug heave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
D.3 Procedure to determine fracture-ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
E Modelling on plug stability 101

E.1 Determination of time lag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
E.2 Schematization of plug uplift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
F Parameters for verification 103

F.1 Case A: Curlew FPSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
F.2 Case B: Gullfaks C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
F.3 Plaxis calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
xiv
List of Figures
1.1 FPSO Aseng prior to operation offshore in Equatorial Guinea, West Africa (SBM
Glossary) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Installation in homogeneous soils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Installation in layered soils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1 Self-weight penetration and suction-assisted penetration during installation . . . . . . 5

2.2 Schematization of suction caisson installation . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Installation of suction caisson with use of WID . . . . . . . . . . . . . . . . . . . . . . 7
2.4 Induced seepage due to cracking (2.4a) or uplift (2.4b) . . . . . . . . . . . . . . . . . . 8
2.5 Installation in layered soils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.6 Normalised contributions of shaft- and tip resistances for effective stress- and CPT
approaches for a typical layered soil profile . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.7 Contributions of plug uplift in homogeneous clay . . . . . . . . . . . . . . . . . . . . . 14
2.8 Equi-potential lines for homogeneous sands (Figure 6 from Senders [2008]) . . . . . . . 15
2.9 Critical suctions for sand overlaid by 3 m clay . . . . . . . . . . . . . . . . . . . . . . . 16
2.10 Comparison of approaches for suction caisson installation of a typical layered soil profile
(this example presents sand overlaid by 3 m of clay) . . . . . . . . . . . . . . . . . . . 18
3.1 Schematization of plug uplift parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.2 Plug failure according to Senders [2008] and Cotter [2009] . . . . . . . . . . . . . . . . 22
3.3 Schematization of plug uplift parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4 Plug failure and bandwidth of cracking failure (clamped and hinged) . . . . . . . . . . 23
3.5 Plug failure for several reference-data (data-labels represent su [kPa]) . . . . . . . . . . 24
3.6 Set-up of tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.7 System-permeability test without clay . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.8 Uplift of clay for relative thick (left series) and thin (right series) clay . . . . . . . . . 29
3.9 Top view of cracks with minor leakage (≈ 1 % of cracks) . . . . . . . . . . . . . . . . . 30
3.10 Top view of cracks for test 3 (left) and 4 (right) . . . . . . . . . . . . . . . . . . . . . . 30
3.11 Reference permeability for test 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.12 Test results of high D/z-ratio compared with model (including bandwith) . . . . . . . 32
3.13 Test-results compared with model, including diameter (D) and thickness of clay plug (z) 33
4.1 stable and moving plug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.2 Suction inside caisson with idealized pressure drop . . . . . . . . . . . . . . . . . . . . 38
4.3 Fracture ratios for high and lower permeable sands . . . . . . . . . . . . . . . . . . . . 40
4.4 Effect of layer thickness on the fracture-ratio requirement . . . . . . . . . . . . . . . . 41
4.5 Volumes during suction-assisted penetration . . . . . . . . . . . . . . . . . . . . . . . . 42
4.6 Volume continuity during uplift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.7 Suction inside caisson with idealized pressure drop . . . . . . . . . . . . . . . . . . . . 43
4.8 Contributions of plug uplift resistances . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
xv
xvi List of figures
4.9 Fracture ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.1 Components of plug uplift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5.2 Pipe/soil interaction after Bridge et al. [2004] . . . . . . . . . . . . . . . . . . . . . . . 48
5.3 Contribution of reverse end bearing for different penetration depths . . . . . . . . . . 49
5.4 Schematization of reverse end bearing during installation . . . . . . . . . . . . . . . . 50
5.5 Time lag for range of permeabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.6 Spring-model for plug uplift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.7 Volume above plug during installation . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.8 Effect of large time-step for calculations . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.9 Dashpot-model for plug uplift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.10 Plug heave due to seepage flow for a range of permeabilities . . . . . . . . . . . . . . . 55
5.11 Combination in series to model plug uplift . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.12 Components of plug uplift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.13 Difference in uplift with penetration depth . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.1 Interpolated cone resistance plot of soil profiles for anchors 1,2 & 3 . . . . . . . . . . . 60
6.2 Predicted suctions for installation of anchors 1 - 3 . . . . . . . . . . . . . . . . . . . . 61
6.3 Measured versus estimated values for installation . . . . . . . . . . . . . . . . . . . . . 61
6.4 Back-calculations with best-fit approach . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.5 Three potential plugs for uplift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.6 Implementation of plug uplift for plug 3 (z = 4.7 m), without contribution of reverse
end bearing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.7 Back-calculations for Case A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.8 Time lage for Case A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.9 Implementation of plug uplift, divided by components . . . . . . . . . . . . . . . . . . 67
6.10 Back-calculations with best-fit approach . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.11 Measured and back-calculated suction (with and without tip reduction) . . . . . . . . 69
6.12 Contributions of uplift components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
6.13 Estimated heave for range of permeabilities . . . . . . . . . . . . . . . . . . . . . . . . 71
6.14 Generated mesh of model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.15 Data-points to investigate behaviour of the plug . . . . . . . . . . . . . . . . . . . . . . 73
6.16 Cases to model cracking of clay plug . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.17 Output for data-points K and L according to Figure 6.15 . . . . . . . . . . . . . . . . 75
6.18 Excess pore pressures for data points D and E . . . . . . . . . . . . . . . . . . . . . . . 75
6.19 Excess pore pressures for data points D and E . . . . . . . . . . . . . . . . . . . . . . . 76
6.20 Calculated plug heave for lower installation-rate . . . . . . . . . . . . . . . . . . . . . . 77
6.21 Calculated plug heave for cracked plugs . . . . . . . . . . . . . . . . . . . . . . . . . . 77
A.1 Estimated inner-, outer, tip- and total resistances according to Houlsby and Byrne
[2005] and Andersen et al. [2008] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
B.1 Bearing capacity factor Nc vs normalised depth zi /D for circular foundation with pure
vertical loading (Fig 4-4 from DnV [1992]) . . . . . . . . . . . . . . . . . . . . . . . . 93
C.1 Schematization of plug cracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
D.1 Illustration of steady-state suction during installation . . . . . . . . . . . . . . . . . . 97

D.2 Soil flow around caisson tip (fig 7.35 of Randolph and Gourvenec [2011]) . . . . . . . . 98
F.1 Parameters for calculations of Curlew case (Case A) . . . . . . . . . . . . . . . . . . . 103

F.2 Soil data of Gullfaks C (Case B) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
List of figures xvii
F.3 Clay content of Gullfaks C (Case B) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

F.4 Material properties of Sand (general) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
F.5 Material properties of Sand (parameters) . . . . . . . . . . . . . . . . . . . . . . . . . 107
F.6 Material properties of Sand (flow parameters) . . . . . . . . . . . . . . . . . . . . . . . 108
F.7 Material properties of Clay (general) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
F.8 Material properties of Clay (parameters) . . . . . . . . . . . . . . . . . . . . . . . . . . 110
F.9 Material properties of Clay (flow parameters) . . . . . . . . . . . . . . . . . . . . . . . 111
F.10 Calculation steps 1 and 2 for initial case . . . . . . . . . . . . . . . . . . . . . . . . . . 112
F.11 Calculation steps 3 - 5 for initial case . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
xviii List of figures
List of Tables
2.1 Overview of effective stress- and CPT approaches for installation in layered soils . . . 10
3.1 Scaling laws for 1G testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.2 Model properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3 Bandwith of pressures below clay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
6.1 Key-data for Case A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

6.2 Soil-data for calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.3 Potential plug dimensions for lift off . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.4 Key-data from Case B anchor installation . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.5 Input-data for Plaxis calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.6 Calculation steps for Plaxis calculations . . . . . . . . . . . . . . . . . . . . . . . . . . 73
xix
xx List of tables
Symbols
Symbol Significance Unit
a = Parameter after method of Andersen et al. [2008] [-]

a1 = Dimensionless pore pressure factor [-]
Abase , Ab , Acan = Cross area of the caisson base [m2 ]
Aclay , Ac = Cross area of clay plug [m2 ]
Ai , Ain = Inner area of the suction caisson [m2 ]
Ao = Outer area of the suction caisson [m2 ]
Atip = Tip area of the suction caisson [m2 ]
c1 = Empirical parameter after Feld [2001], here 0.36 [-]
c2 = Empirical parameter after Feld [2001], here 0.48 [-]
co = Empirical parameter after Feld [2001], here 0.45 [-]
D = Diameter of suction caisson [m]
Di = Inner diameter of suction caisson [m]
Do = Outer diameter of suction caisson [m]
fskirt = Total skirt resistance [kPa]
h = Penetration depth [m]
hc = Penetration in the clay [m]
hs = Penetration in the sand [m]
hwt = Height of water table [m]
icrit = Critical hydraulic gradient [kPa]
K = Lateral earth coefficient for penetration of skirt [-]
kclay = Permeability of clay [m/s]
kcrit = Critical permeability [m/s]
kf = Friction coefficient [-]
kf = Permeability factor after Andersen et al. [2008] [-]
Kf = Lateral earth pressure coefficient [-]
ki = Permeability inside suction caisson [m/s]
ko = Permeability outside suction caisson [m/s]
kp = End-bearing coefficient [-]
ksand = Permeability of sand [m/s]
Kt = Tip resistance factor [-]
Nc = Bearing capacity factor to account for cohesion [-]
Nq = Bearing capacity factor to account for surcharge [-]
Nγ = Bearing capacity factor to account for self-weight [-]
pcrack = Pressure for plug cracking [kPa]
pplug = Pressure for plug uplift [kPa]
Pbearing = Force due to reverse end bearing [kN]
Pf riction = Force due to friction of clay plug [kN]
PREB = Reverse end bearing capacity [kPa]
Ptotal = Sum of Pweight , Pf riction and Pbearing [kN]
Pweight = Submerged weight of the clay plug [kN]
qc = Measured cone resistance [kPa]
qc,clay = Measured cone resistance in clay [kPa]
qc,sand = Measured cone resistance in sand [kPa]
xxi
xxii Symbols
Symbol Explanation Unit
Qclay = Flow through clay [m3 /s]

Qcrit = Critical flow [m3 /s]
Qinside = Inside resistance of skirt [kN]
Qoutside = Outside resistance of skirt [kN]
qplug = Specific flow in plug [m/s]
Qs,max = Maximum uplift force after Bridge et al. [2004] [kN]
Qsand = Flow through sand [m3 /s]
qsand = Specific flow in sand [m/s]
Qskirt = Total skirt resistance (inner and outer) [kN]
Qtip = Tip penetration resistance [kN]
qtip = Tip resistance [kPa]
Qtot = Total penetration resistance [kN]
r = Radius of suction caisson [m]
r = Roughness factor after Houlsby and Byrne [2005] [-]
s = Seepage length in sand [m]
Scrit = Critical suction [kPa]
Scrit,clay , Splug , Pplug = Critical suction in clay [kPa]
Scrit,N GI = Critical suction after Andersen et al. [2008] [kPa]
Scrit,sand = Critical suction in sand [kPa]
splug = Incremental uplift of the plug [m]
splug,seepage = Uplift of the plug by seepage [m]
splug,suction = Uplift of the plug by applied suction [m]
Sred , S2 , ∆S2 = Reduced suction, just below clay plug [kPa]
Sreq = Required suction for installation of caisson [kPa]
su = Undrained shear strength [kPa]
t = Thickness of suction caisson [m]
T = Time lag [s]
Tmax,clamped = Maximum circular bending moment (clamped) [kNm]
Tmax,clay = Maximum soil resistance of clay [kNm]
Tmax,hinged = Maximum circular bending moment (hinged) [kNm]
v = Installation velocity [m/s]
V = Volume [m3 ]
Vbox = Volume in soil-box [m3 ]
vcan = Velocity of suction caisson [m/s]
vplug = Velocity of plug [m/s]
Vwc = Volume in water column [m3 ]
W = Work generated by underpressure [kNm]
W’ = Submerged weight of suction caisson [kN]
z = Penetration into soil [m]
zclay = Thickness of clay plug [m]
α = Dimensionless adhesion factor [-]
αin = Factor for reduction of outside resistance [-]
αout = Factor for reduction of inside resistance [-]
αtip = Factor for reduction of tip resistance [-]
γc0 = Effective volumetric weight of clay [kN/m3 ]
γs0 = Effective volumetric weight of sand [kN/m3 ]
γw = Volumetric weight of water [kN/m3 ]
∆hwt = Differential height of water table [m]
∆t = Time-step [s]
∆Sdispl,plug = Differential volume below clay plug [m3 ]
∆Vdisplaced , ∆Vdisp ,
∆Vdispl,can = Displaced volume due to penetration of caisson [m3 ]
∆Vpump = Pumped volume out of caisson [m3 ]
∆Vseep,clay = Volume due to seepage through clay [m3 ]
∆Vseep,sand = Volume due to seepage through sand [m3 ]
∆Vseepage = Volume due to seepage inflow [m3 ]
ν = Poisson’s ratio [-]
σv0 = Vertical effective soil stress [kPa]
0
σvi = Vertical inner effective soil stress [kPa]
0
σvo = Vertical outer effective soil stress [kPa]
φe = External friction angle [◦ ]
φi ’, φ’ = Internal friction angle [◦ ]
Chapter 1
Introduction
1.1 Background information

SBM Offshore (SBM) is a leading offshore engineering company in floating production and moor-
ing systems. As part of the mooring systems suction caissons are frequently used as anchors for
single-point-mooring systems and spread moored systems. These suction caissons are large diameters
steel piles and installed by creating an underpressure inside the pile by suction. The underpressure
generates a net downward force which contributes to the penetration of the caisson for either homo-
geneous clayey or sandy soils. For the latter condition, seepage flow is induced which contributes to
tip resistance reduction of the caisson.
The main advantages of suction caissons compared to conventional offshore anchors are the relat-
ively quick installation time and the possibility to remove the caisson by the reverse principle. Usually
less equipment is required and smaller extraction spreads are mobilized for installation. Nowadays
suction caissons are extensively used for FPSOs (Floating Production, Storage and Offloading) as
presented in Figure 1.1 (Tjelta [2001]).
Figure 1.1 FPSO Aseng prior to operation offshore in Equatorial Guinea, West Africa (SBM Glossary)
1.2 Problem description

In general the installation of a suction caisson consist of two phases; initial self-weight penetration
(SWP), followed by a suction-assisted penetration (SAP). The initial self-weight penetration depends
on the effective weight of the suction caisson in relation to the soil resistance. Ballast weight can
be used to increase the SWP. The second phase consists of a suction-assisted phase, which initiates
1
2 Installation effects of suction caissons in non-standard soil conditions
with a transitional phase where the pumping is started to remove the trapped water. Installation of
suction caissons in coarse grained soils (high permeable soils) is achieved by the induced seepage flow
through the soil (Figure 1.2a). This reduces the soil resistance at the caisson tip, which contributes to
the installation process. The upward groundwater flow also reduces effective soil stresses and hence
reduces the horizontal stress against the skirt (Erbrich and Tjelta [1999]).
Water out Water out
Underpressure Underpressure
Sand Clay
Seepage flow
(a) Installation in sand (b) Installation in clay
Figure 1.2 Installation in homogeneous soils
For installation in clays (low permeable soils), the suction process is not creating significant
seepage flow because of the lower permeability. In this case, the installation is achieved by the
pressure difference over the top cap, i.e. the downward gradient of the water column provides the
force to push in the caisson into the clay (Figure 1.2b).
Installation in layered soils, especially where sand overlaid by clay, is expected to be more chal-
lenging since the potential reduction of tip resistance is somehow restrained from flow restrictions
(Figure 1.3a). It is not clear to what extent the lower permeable layer will impact the seepage and
thus the tip reduction, which possibly is required to reach target detph.
Water out Water out
Clay
Clay
Sand Sand
Seepage flow
(a) Restriction of flow in underlying sand (b) Plug uplift and seepage flow
Figure 1.3 Installation in layered soils
The stability of the plug becomes an important aspect once the seepage flow is required to reduce
the tip resistance and hence to install the caisson to target depth. When uplift of the clay plug occurs
(Figure 1.3b) it is uncertain to what extent this generates a seepage flow. One of the main concerns
is excessive plug uplift (i.e. when the plug contacts the top cap), which could occur when higher
suction pressures are applied or maintained for a longer period than predicted. Plug uplift probably
creates seepage flow within the underlying sand and thus reduces the tip resistance to some extent.
R.H. Romp M.Sc. Thesis

Introduction 3
On the other hand, plug uplift will become problematic once the plug contacts the top cap and no
suction (and thus no downward force) can be generated anymore.
This thesis focusses on the aspects of suction caisson installation in non-standard soil conditions,
especially where sand is overlaid by clay. Installation aspects of suction caissons are extensively de-
scribed by many authors for homogeneous clay (e.g. Houlsby and Byrne [2005]) or granular material
(e.g. Andersen et al. [2008]). However, for installation aspects in layered soils limited documenta-
tion, references or guidelines are available (Senders [2008] and Cotter [2009]). Furthermore, at present
safety measures are implemented in the design phase to overcome the uncertainties of excessive plug
heave.
These uncertainties incorporate conservatisms, which could have a great impact on the feasibility
of the suction caisson concept. A detailed research of the plug stability aspects will contribute to a
better understanding of the governing aspects involved and hence reduce certain uncertainties at the
design phase.
1.3 Purpose of research

The main purpose of this research is to increase understanding and gain insight into the governing
mechanisms during the installation of a suction caisson in conditions where sand is overlaid by clay.
Within this framework, the following objectives are formulated:
- Literature study on installation aspects of suction caissons in both homogeneous- and layered
soils (sand, clay and/or silt strata);
- Formulation of an analytical model to describe the governing mechanisms associated with the
installation of the caisson, i.e. cracking of the plug, variable installation-rate, plug heave;
- Gain insight in the mechanism associated with plug stability using basic experiments conducted
at the Geotechnical laboratory of Delft University of Technology;
- Back-calculation of two representative field cases using the developed analytical model;
- Comparison of the analytical model with a commercially available geotechnical finite-element

code (PLAXIS).
The research question to be answered is based on the objectives, available time and budget. The
limitations of this research are as follows:
- Only installation aspects will be investigated (i.e. the holding capacity is not considered);
This represents the penetration of the suction caisson until the design depth is reached. Neither
influences of in-place conditions nor removal aspects of the suction caisson will be investigated.
As well, no time-dependent effects during the lifetime of suction caissons are taken into account
(e.g. set-up effects, creep in the subsoil, erosion).
- Only installation in layered soils is considered;

The literature review discusses installation aspects of suction caissons in layered soil (i.e. sand,
silt and/or clay strata). Further assessment of installation involves installation of sand profiles
(silica sands) overlaid by clay.
- A simplified geometry for the suction caisson is considered;

No effects of ring stiffeners or pad-eye stiffening will be taken into account wihtin the analyses.
Structural influences of buckling and/or radial expansion/compression will not be taken into
account. Soil layers are assumed to be horizontally deposited and suction caissons penetrate
vertically in the soil. Small deviations from these idealized conditions are not considered in this
research.
M.Sc. Thesis R.H. Romp

1.4 Readers manual

This document includes the methods to calculate the resistances for installation of suction caissons
in layered soil (Chapter 2). The next Chapter (3) discusses the plug response during installation of
a caisson in layered soils. Chapter 4 focuses on the plug stability aspects, followed by a method to
model the components of plug uplift in Chapter 5. Hereafter verification of the theory is presented
by back-calculations of two cases and numerical calculations in Chapter 6. The conclusions and
recommendations are presented in Chapter 7.

Chapter 2
Installation in layered soils
This Chapter discusses the geotechnical engineering aspects of a suction caisson installation in con-
ditions where sand is overlaid by clay. First the general aspects of suction caisson installation are
discussed in section 2.1, such as tip-reduction and seepage flow for installations in sandy soils. As
well the effect of penetration-rate is presented.
In the following section (2.2) a description of the self-weight installation in a layered stratum is
discussed considering conventional effective stress- and CPT-approaches as presented by the stand-
ards (DnV [1992]; API [2000]). These approaches are adjusted for layered soil conditions, which is
not included in the standards. Once self-weight penetration has been completed, suction assisted
penetration is applied. This condition is further discussed in section 2.3, where again the effective
stress- and CPT-approaches are considered, together with the empirical approach of Senders [2008].
2.1 Installation of a suction caisson

In general the installation of a suction caisson consist of two phases; initial self-weight penetration,
followed by a suction-assisted penetration (Figure 2.1). The initial self-weight penetration clearly
depends on the effective weight of the suction caisson but can be increased by an additional weight
on top of the caisson during penetration. In any case, it is recommended to ensure sufficient initial
penetration prior to the suction assisted phase. For sands it holds that initial self-weight penetration
should be deep enough to avoid piping effects. In case of homogeneous clays, the sealing is of great
importance. From several practical cases it can be found that around 1 m of initial penetration is
sufficient (Tjelta et al. [1986]).
Lower caisson + suction pump Caisson touchdown & initial

self-weight penetration
Pump water out caisson untill target Undock of suction pump

depth is reached
Figure 2.1 Self-weight penetration and suction-assisted penetration during installation
5
Installation in sand
For sands the second phase of installation consists of; first a transitional phase followed by a suction-
assisted phase. During the transitional phase, the pumping is started to pump out the trapped
water. As a result, the relative pressure inside the caisson is lowered, which attracts the underlying
pore water to flow inside the caisson in case of sand strata (Figure 2.2a). Around the skirt tip the
induced water flow reduces the effective stress locally, which contributes to the penetration of the
caisson. During the suction-assisted phase the pressure inside the caisson drops gradually, due to the
combined effect of attracting pore water and pumping out water the caisson. In the next phase the
rate of attracted pore water becomes constant, as well as the pumping-out rate. This phase remains
constant until target penetration depth has been reached.
Installation in clay
In case of homogeneous clay, the relative pressure inside the caisson is lowered as well. In comparison
to sand, it can be remarked that the permeability of the clay is much lower and therefore no seepage
flow will be induced during a typical installation period (Figure 2.2b). Installation of caissons in clays
is achieved by displacing the trapped water column, which provides the force to push the caisson into
the clay. Apart from the seepage flow, this principle holds as well for sands and becomes very effective
for deep-sea installations.
Water out Water out
Clay
Sand
Seepage flow
(a) Installation in permeable soil (b) Installation in low permeable soil

Sand
Figure 2.2 Schematization of suction caisson installation
2.1.1 Reduction of tip resistance in layered soils

Common installation of suction caissons in sandy soil is based on the principle of tip reduction due
to seepage flow, induced by the applied suction. For installation in clays, the tip reduction is not
applicable due to the impermeable soil. The soil resistance of clays in general is lower compared to
caisson installations in sands. As mentioned in Chapter 1, installation in sand overlaid by clay could
be problematic since the seepage is restricted by the clay layer.
However, according to several researchers (Tran et al. [2007], Senders [2008] and Cotter [2009]),
it has been found that some reduction in underlying sand was monitored for installation in layered
soil. A well-known full-scale experiment of installation in layered soil was performed by Tjelta et al.
[1986] using a large diameter trial test. As well the installation of 9 suction caissons for the Curlew
FPSO (Alhayari et al. [1999]) and installation of 12 suction caissons for the GORM field (Senpere
and Auvergne [1982]) were successful. All installations were performed in layered soil conditions
(either stratified clay/silt and sand) and successful installation could have been achieved by the use

Installation in layered soils 7
of water-injection devices near the caisson tip. Application of these devices contributes to reduction
of the effective stresses, by injecting water in front of the caisson tip. As a result the tip-resistance
is ’artificially’ decreased, which contributes to succesfull installation in complex soil conditions.
Water injection devices
Usage of water-injection-devices (WID) seems to be an efficient mitigating measure to install suction

caissons in complex soil conditions. However, this measure consists of both water injection and
extraction (suction), which is not efficient from an engineering point of view (Figure 2.3). The
effectiveness of WIDs is nowadays not well understood, what raises the question to the use of this
measure. During installation with WIDs the soil around the skirt tip is remoulded along the skirt.
When high pressures are applied (jetting) the soil is flushed and the effective stresses reduce to zero.
When pressures are more gradual (injection) the effective stresses reduce to a lesser extent (Cotter
[2009]). Both conditions could influence the holding capacity of the caisson during lifetime in a
negative way. However, this aspect is not taken into account within this thesis. Quantification of
the effectiveness is desired, which in this thesis will be assessed by considering the plug stability
during installation. This thesis only focuses on the installation effects of suction caissons without
any contribution of WIDs.
Water out Water out
Water out
Underpressure
Clay Clay
Clay
Sand Sand
Sand
(a) Soil resistance too high (b) WID switched on;

(c) Penetration of caisson
for penetration reduction of tip resistance
Figure 2.3 Installation of suction caisson with use of WID
2.1.2 Seepage flow in sand

In case of homogeneous coarse grained soils; i.e. sands, the installation of a suction caisson is achieved
by the reduction of the tip resistance. At the tip of the skirt the flow gradient will be high due to
piping-effects. As a result the grains will (partly) liquefy around the tip and reduce the effective
stresses. This mechanism of partly liquefaction will occur on a very small scale around the caisson
tip. Because the induced flow is towards the inside of the suction caisson, the liquefied grains will
flow inside the caisson, which result in some plug heave after installation. The theoretical minimum
heave to be expected is equal to the volume of the installed skirt (Appendix D.2).
The required installation forces in layered soil are lower than for jacked installation according to
several authors (Watson et al. [2006], Senders [2008] and Cotter [2009]). Experiments of Watson et al.
[2006] show that installation of suction caissons in layered soil still can be achieved in an economical
way. This was concluded from the fact that the required suction force was less than predicted for
jacked installation. As result of this, reduction due to seepage flow is one of the possible mechanisms
that caused the lower installation resistances. This could imply that a reduction of the tip resistance

and thus seepage flow did occur during installation. Senders [2008] describes this principle in detail,
whereas in general two mechanisms are identified which could have occurred (Figure 2.4).
Water out Water out
Clay
Clay
Sand Sand
Seepage flow Seepage flow
(a) Mechanism I: Cracking of the clay plug (b) Mechanism II: Uplift of the clay plug
Figure 2.4 Induced seepage due to cracking (2.4a) or uplift (2.4b)
Mechanism I: Mechanism II:

Cracking of clay layer, which induces the seep- Uplift of the entire (intact) inner clay plug.
age flow to develop through the underlying This uplift transforms the under pressure in
sand. Depending on the soil properties crack- the trapped gap to just below the sealing clay
ing is likely to occur in relatively thin top lay- layer. This way seepage flow is induced by the
ers, in relation to the caisson diameter (see transferred ’suction’ pressure and some reduc-
Figure 2.4a). tion of tip resistance will occur
Senders [2008] found that during the beginning of the suction installation the top layer can be
disturbed or fractured (mechanism I). The seepage in the sand layer can thereafter occur similar as
for the uniform sand mechanism. This effect is expected to occur for relative thin clay overlying sand
profiles (see Figure 2.4a).
Especially in case of mechanism II it was found that the uplift is not dependent on the stratific-
ation of the soil, e.g. uplift occurs as well for clay over sand profiles as for interbedded clay in sand
strata. According to Tran [2005] the influence of stratification has been investigated for silt layers
with permeability roughly 2 orders of magnitude lower than for the underlying sand that was used.
Variable top layer thicknesses and interbedded thicknesses were investigated, for the non-plastic silty
material. From the results, it showed a rapid drop in suction pressure for silt interbedded in sand
layers. Interesting is that the same effect occurs for silt at depth 0,8 m as well as for 2,0 m below
surface. This implies that no depth-dependent effect was encountered.
Cotter [2009] investigated the installation process for inclined clay profiles, which results in gradually
seepage restrictions. However, as presented by many other authors (Tran et al. [2007], Senders [2008]
and Senpere and Auvergne [1982]), the installation was still successful.
From the work of several authors investigated, it can be remarked that the stratification is related to
the mechanism to be expected. However, this only is valid if the thickness of the layer is considered,
e.g. relative thick impermeable (clay) layers have the tendency to lift up. Considering the location
of flow-restricted layers, there is no depth-dependant effect encountered.

2.1.3 Effects of penetration-rate

Regarding the installation aspects, several researchers found that possible plug uplift is a function
of the installation-time (Senders [2008]). The effect of elastic plug heave (presented in Appendix
D.2) is not rate-dependent and is excluded for rate-dependency effects. Fast suction installation
implies a greater pressure difference of inner- and outer pressure, which influences the plug response
during installation. Cotter [2009] found that for lower suction pressured installation, plug heave
was observed in all cases. Watson et al. [2006] observed that plug uplift accelerates towards the
final stages of penetration, which also implies that the magnitude of plug uplift is dependent on the
applied suction.
The authors describe the uplift mechanism being dependant on the installation time, e.g. short
installation time implies lower plug uplift. Mechanism I (cracking of the plug) is not considered
in relation to the installation-time. It is expected that erosion will determine the time-dependent
behaviour for this condition. From practical experiences it has been observed that longer installation
time involves higher risks. In order to control the risks involved during installation (e.g. installation
refusal), a high flow rate for the pumps is therefore desired.
2.1.4 Conclusions
From practical experiences about installation of suction caissons in layered soils, it was found that
succesfull installation is achieved by the use of water-injection-devices. The theory about installation
in sand overlaid by clay identified two mechanisms; cracking- or uplift of the clay plug. From both
practical experiences by several researchers it was concluded that fast installation limits the plug
uplift and is therefore desired to control the risks involved, such as installation refusal.
2.2 Initial phase: Self-weight penetration

This section presents the self-weight penetration phase for a suction caisson installation in layered
soils. First homogeneous sand and clayey soils are considered by available prediction methods. The
self-weight penetration depth (h) can be calculated by the resistance of skirt skin friction, end bearing
and submerged weight of the caisson. The latter acts as the driving force for initial penetration. The
depth of penetration can be calculated by solving the equation, which is dependent on the unknown
parameter h. The solution for this parameter equals the downward forces (submerged weight) with
the encountered resistances (skirt skin friction and tip end bearing).
If the downward force of self-weight is higher than the sum of resistance forces, the caisson will
penetrate. The existing procedures for predictions of penetration resistances can be divided into the
following two categories;
Effective stress-approach; based on tip- and shaft resistance in proportion to the effective
stresses.
Houlsby and Byrne [2005] and API [2000] provide common methods which calculate the resistances
based on effective stresses. Only the Houlsby and Byrne-method presents an approach to predict the
soil resistances in the suction-assisted phase. These calculations are rather extensive and therefore a
simplified Houlsby and Byrne-method is adopted to calculate the soil resistances during penetration.
CPT-approach; based on tip and shaft resistance in proportion to cone resistance qc .
The DnV [1992] method is the most conventional method which calculates the soil resistances during
self-weight penetration. However, the potential reduction during suction-assisted penetration is not

included. Andersen et al. [2008] presented a procedure to calculate the reduction considering seepage
flow for dense sands, which also uses the CPT-data for predictions. A combination of both methods
is used to describe the installation process of the suction caisson. In addition Senders [2008] described
an approach for layered soil profiles, especially where sand is overlaid by clay.
Table 2.1 presents an overview of the methods and their applicability. A simplified Houlsby &
Byrne-method as effective stress-approach and the DnV-method as CPT-approach are discussed in
the next paragraphs.
Table 2.1 Overview of effective stress- and CPT approaches for installation in layered soils
Approach Self-weight Suction- Sand/Clay/ Layered CPT/σv0 /

penetration assisted Both soils Both
penetration
Houlsby and Byrne [2005] yes yes Sand + Clay no σv0
API [2000] yes no Sand + Clay no CPT
DnV [1992] yes no Sand + Clay no CPT
Andersen et al. [2008] yes yes Sand no Both
Senders [2008] yes yes Both yes Both
Simplified H & B yes yes Both yes σv0
In case of layered soils the self-weight penetration depth can be calculated by extrapolating this
principle for multiple layers. However, a transition in friction and end bearing resistance shall be
considered for penetration through the interface of different layers.
2.2.1 Effective stress-approach

As stated previously, the Houlsby & Byrne-method is rather extensive to predict the soil resistances
during penetration. The self-weight penetration phase of the suction caisson in the clay layer is
calculated as the sum of skirt friction and the end bearing on the tip (see Figure 2.5).
Qtot = Qinside + Qoutside + Qtip (2.1)
If the undrained shear strength of the clay is considered to be uniform and adhesion factors of
inside and outside are the same, the equation becomes less extensive (equation 2.2).
Qtot
Qtot
Suction caisson Suction caisson

Qsuction
hc Qinside Clay Qoutside Clay

h hc Qinside,clay Qoutside,clay
Qtip Qtip hs Qinside,sand Qoutside,sand
Sand Qtip
Sand Qtip
(a) Penetration in clay layer (b) Penetration in underlying sand layer
Figure 2.5 Installation in layered soils
The self-weight penetration depth in the clay (hc ) can be calculated by assuming the submerged

weight (W’ ) equal to the penetration resistance (Qtot ). Undrained behaviour is applicable since no
significant consolidation will occur during a typical installation time period.
W 0 = (Ai + Ao )αsu hc + (su Nc + γc0 hc )Atip (2.2)
From equation 2.2 it can be noted that the dimensionless adhesion factor is taken equal for inside
and outside of the caisson in order to simplify the equation. De parameters Ai and Ao are defined
per unit meter of penetration. The end bearing parameter is based on DnV [1992] and equal to 6.2
for shallow penetration and will increase up to 9 for deeper levels of penetrations (see Appendix B).
If the submerged weight exceeds the soil resistance in the clay, the penetration will continue in
the underlying sand (see Figure 2.5b). The self-weight penetration phase in sand can be calculated
by the same approach, i.e. the penetration resistance consists of inner and outer skirt friction plus
end bearing at the tip. However, drained behaviour is considered for the sandy material. The skirt
friction can be calculated by a lateral earth pressure coefficient of K = 0.8, which is used for the
prototype cases in Andersen et al. [2008]. The value of K = 0.8 is also recommended by API [2000]
for the calculation of the drained shaft friction of open-ended unplugged piles.
A simplification of this method (Simplified H & B ) ignores the difference in vertical effective
stress for inside and outside the caisson, due to differences in area of influence (σv0 = σvi
0 0
= σvo , from
Houlsby and Byrne [2005]). The angle of friction is estimated by φe =0.9 φi , where the internal
friction angle is based on triaxial tests and drained conditions are considered for installation in sand.
The resistances are calculated by;
1
Qinside = Qoutside = Qskirt = πDKtan(φe )σv0 hs (2.3)
2
t
Qtip = (σv0 Nq + γs0 Nγ )Atip (2.4)
2
1 t
W 0 = (Di + Do ) πKtan(φe )σv0 hs + (σv0 Nq + γs0 Nγ )Atip (2.5)
2 2
2.2.2 CPT-approach
According to Senders and Randolph [2009] the cone penetration resistance is closely related to the
measured resistance during self-weight penetration of a suction caisson. This relation relies on the
similarities in penetration speed and dimension of the cone compared to the thickness of the caisson.
Since the shape of the cone is different compared with the strip-shape of the caisson skirt, shape
factors are introduced. This shape factor relates the cone resistance to the caisson resistance. For
this reason the CPT-based method is considered to be reliable to predict self-weight penetration.
The principle of this method is given by the DnV [1992], which is defined by;
Qtot = Qinside + Qoutside + Qtip (2.6)
Z L
Qoutside = πDo kf qc (z)dz (2.7)
0

Z L
Qinside = πDi kf qc (z)dz (2.8)
0
Qtip = Atip kp qc (L) (2.9)
Where the coefficient kf varies between 0.001 (most probable) and 0.003 (highest expected), for
kp a factor of 0.3 to 0.6 is suggested. These values are based on North Sea silica sand, which is typical
dense sand.
2.2.3 Penetration in layered soils

The self-weight penetration in layered soil can be calculated using both approaches. In case of the
effective stress-approach, the equations are adjusted to estimate the total resistance in layered soils.
The calculation principle of soil resistance in homogeneous sand is extended by the friction of the
upper clay layer, if is assumed that penetration through the top clay layer occurs.
1 t
W 0 = (αo Ao + αi Ai )su hc + (Do + Di ) πKtan(φe )σv0 hs + (su Nc + σv0 Nq + γs0 Nγ )Atip (2.10)
2 2
The total penetration depth (h) will be the depending variable to be calculated and is defined by;
h = hc + hs (according to Figure 2.5b) . A distinction between the penetration in clay or underlying
sand can made, with the end bearing conditions;
h ≤ hc : Nc = 6.2 - 9 Nq = 1 Nγ = 0
h > hc : Nc = 0 Nq > 1 Nγ > 0
In case of the CPT-approach, the penetration resistance can directly be related to the measured
cone resistance. This is advantageous from an engineering point of view, since the measured soil
resistance is related to a single factor, i.e. kp and kf . Although these factors are empirical factors
and site dependent, this is considered to be the most robust method.
2.2.4 Comparison of approaches

The differences between both approaches generally rely on the difference in input parameters. The
effective stress-approach uses the effective soil stress and friction angle to calculate the soil resistance,
by considering drained behaviour during installation. The CPT-approach directly relates the cone
resistance to the predicted soil resistance. Disregarding the limited time consumption for interpreta-
tion and soil data needed, the main advantage of the CPT-approach is based on the penetration-rate.
Similar soil response is obtained in terms of (un-)drained or partially drained behaviour. Addition-
ally, limiting input parameters are required to predict the soil resistance for tip and skin friction
(kp or kf ).
In order to compare both approaches properly, similar input conditions are required. Thus the
in-situ soil strength should be correlated to CPT-data, which is done according to Verruijt and van
Baars [2007] in equation 2.11 and Chen and Juang [1996] in equation 2.12;
qc,clay = σv0 + Nc su (2.11)

0
qc,sand = σv0 0.266e6.82tan(φ ) (2.12)
These equations are used to compare the effective stress-approach and CPT-approach for a typical
case. The ratios of friction and end bearing relative to the total resistance for increasing depth are
plotted in Figure 2.6, where sand with φ0 = 40◦ is overlaid by 3 m of clay with su = 25 kPa (uniform
with depth). For the CPT-approach the kf = 0.0015, kp = 0.45 (most probable) as best-fit and
kf = 0.003, kp = 0.6 (highest expected) are taken according to Senders and Randolph [2009]. The
Nc parameter for the clay is calculated according to DnV [1992] (see Appendix B). From Figure
2.6 it can be seen that the used correlations agree reasonably well, for the two different approaches.
Further details of the calculations are presented in Appendix A.3.
Q_fric or Q_tip [-]

Q_tot Q_tot
0,00 0,20 0,40 0,60 0,80 1,00
0,00
1,00
2,00 Clay
3,00
4,00
Depth (m)
5,00 Sand Shaft
6,00
Houlsby & Byrne Tip
7,00
NGI (most
8,00
probable)
9,00 NGI (highest
expected)
10,00
Figure 2.6 Normalised contributions of shaft- and tip resistances for effective stress- and CPT approaches
for a typical layered soil profile
It can be noted in Figure 2.6 that at the initial stage of penetration (i.e. within the clay layer)
the total resistance is governed by the shaft resistance (shaft). With increasing penetration in the
underlying sand it can be seen that end bearing becomes the governing resistance (tip). Tip reduction
due to seepage flow would then become an effective measure to reduce the total soil resistance in the
underlying sand. Further discussion about seepage flow in the underlying sand is presented in the
subsequent section (2.3) and Chapter 4.
2.2.5 Conclusions
In order to predict the soil resistance during the self-weight penetration phase limit state methods are
applied. In general there are two options; effective stress- or CPT-approach. The difference between
the methods relies on key-input parameters, such as friction angle and effective unit weight (effective
stress-approach) or cone resistance and shape-factors (CPT-approach).
The Houlsby & Byrne method is based on effective stresses, which are calculated iteratively. In
addition, this describes the stress changes (reductions) for suction-assisted penetration, which makes
this approach useful. On the other hand the reduction of effective stresses are not applicable for
layered soil conditions, thus some adjustments should be made. This is done by introducting a sim-
plified Houlsby & Byrne method which is less extensive.

From an engineering point of view, the CPT-approach is preferred because the relation to the meas-
ured resistance is directly related by a single factor. As well the CPT should encounter the same soil
response as for the caisson during installation. Considering the penetration-rate, similar soil response
is obtained in terms of (un-)drained or partially drained behaviour. Only the shape factor is different
and should therefore be chosen properly. According to Senders [2008] and Andersen et al. [2008] it
is possible to estimate the reduced stresses during the suction phase.
2.3 Second phase: Suction-assisted penetration

The suction-assisted penetration phase can be characterized by considering the amount of reduction
in soil resistance from seepage flow in the underlying sand layer. In order to predict the tip (and
inner-) reduction, an upper boundary for the applied underpressure is introduced, known as the
’critical suction’. This principle is explained below for clay, sand and layered soil profiles. Hereafter
the reduction in soil resistance is assessed by the effective stress- and CPT-approaches.
2.3.1 Critical suction in layered soils

As the suction below the caisson lid increases, the pressure difference on the soil plug will increase.
This gradient is limited to the ’critical suction’, from which the effective stress becomes zero. For
suction pressures higher than the critical suction, liquefaction is expected for sandy soils, while
for homogeneous clays intact uplift of the plug ultimately occur. The critical suction represents the
maximum suction that can be applied to ensure plug stability. This critical suction is well documented
for homogeneous clays (Houlsby and Byrne [2005]) or sands (Andersen et al. [2008]; Senders [2008]),
however not extensively for layered soils. Because of this, the critical suction is determined on the
basis of relations for the critical suction for clay and sand. By combining these calculations, an
expression for layered soils is obtained.
In clay
In homogeneous clay the critical suction is based on the reverse end bearing (Pbearing ), shaft resistance
(Pf riction ) and submerged weight of the clay plug (Pweight ). Reverse end bearing is defined as the
mobilized uplift resistance from end bearing in the short to medium term (Randolph and Gourvenec
[2011]). Figure 2.7 presents the contributions of each component for a caisson with D = 10 m and
αsu = 8 kPa. The parameter z represents the thickness of the clay plug.
Pfrac / Ptotal
0 0,2 0,4 0,6 0,8 1
0,0
0,1 Pweight /
Ptotal
0,2
Pfriction /
0,3 Ptotal
0,4 Pbearing /
z/D [-]
Ptotal
0,5
0,6
0,7
0,8
0,9
1,0
Figure 2.7 Contributions of plug uplift in homogeneous clay

In absolute terms the contribution of reverse end bearing is independent of the penetration depth
for constant su (uniform profile). Considering the total resistance for plug uplift, it can be seen
that for increasing depth the relative contribution of reverse end bearing decreases for increasing
penetration in the clay, by assuming a constant Nc -factor.
For layered soil profiles, it is assumed that the self-weight installation completely penetrated
through the top clay layer and the contribution of reverse end bearing capacity cannot be taken into
account (Senders [2008]). From Figure 2.7 it can be found that the frictional component will be lower
than the self-weight component in that case. The self-weight component is not depedent on caisson
4γ 0
dimensions and will be the governing component for uplift when Dαscu >> 1. The theory of Senders
[2008] assumes effective unit weight of the clay plug, by considering low uplift rates and hence the
submerged weight of the plug can be taken. The critical pressure for the clay plug is defined by the
uplift criteria in equation 2.13;
Di
Scrit,clay = γc0 zclay + αsu zclay (2.13)
4
In sand
In case of homogeneous sand, the critical pressure is found by the ’exit’ gradient and seepage length.
The limiting factor is the gradient across the upper surface of the plug (bed level), despite higher
hydraulic gradients are expected at the pile tip. However, the soil around the pile tip is confined by
non-liquefied soil and thus liquefaction will intervene first at bed level. Therefore the gradient at pile
tip level is not the limiting factor, but the exit-gradient at bed level. The critical pressure depends
on the effective weight of the soil and empirical relations for the seepage length;
Scrit,sand = sγw icrit = γs0 s (2.14)
(a) Homogeneous sand (b) Sand overlaid by impermeable layer
Figure 2.8 Equi-potential lines for homogeneous sands (Figure 6 from Senders [2008])
For the underlying sand in a layered soil profile, this seepage length is described by Senders [2008].
Figure 2.8 presents the development of pore pressures when suction is applied in the caisson. For
layered soil conditions the seepage length will increase due to the impermeable top layer of clay as
can be noted in Figure 2.8. Senders [2008] presented an empirical correlation for the seepage length,
based on the condition with an impermeable top layer by;
s h −0.85
s
= 1 + 0.3 (2.15)
hs D

If equations 2.14 and 2.15 are combined, an expression for the critical pressure is given by;
h −0.85
s
Scrit,sand = γs0 hs 1 + 0.3 (2.16)
D
The methods by Andersen et al. [2008] and Houlsby and Byrne [2005] also use a critical suction
pressure for homogeneous sands. The normalized critical suction according to the researchers for
homogeneous sands is presented in Figure 2.9. The example presents a 3 m clay layer overlying sand
at (z/D = 0.3). It can be remarked that for penetration in the sand, the critical suction estimate
for homogeneous sand (dashed lines) tend to zero. This is due to ’shallow’ penetration in the sand
and in theory is more favourable to piping failure. However, since the sand is overlaid by clay,
piping failure is less likely to occur. A correction for the Houlsby and Andersen-methods is therefore
desired, which accounts for the depth effects of penetration. This is done by the author by taking the
Senders-solution for low penetration as a starting point. In fact the Houlsby and Andersen-methods
are shifted horizontally towards the Senders-solution for the initial penetration in the sand.
Scrit/y'z
0,0 0,5 1,0 1,5 2,0
0
Senders (2008) Layered
0,1 Houlsby (2005) Sand
Andersen (2008) Sand
0,2 Corrected Houlsby (2005) Layered
Corrected Andersen (2008) Layered
0,3
0,4
z/D
0,5
0,6
0,7
0,8
0,9
Figure 2.9 Critical suctions for sand overlaid by 3 m clay
From Figures 2.8 and 2.9 it can be noted that the critical suction for layered soils is higher
than homogeneous sands, due to the increased seepage length. It can also be remarked that the
corrections for critical suction of the Houlsby and Andersen-methods show a close relation with the
Senders-method for layered soils.
2.3.2 Reduction of soil resistance

In order to ensure tip reduction at the caisson tip a seepage flow should be induced. This can be
achieved by applying suction within the caisson, which will eventually induce a seepage flow (due
to plug cracking or uplift). The seepage flow around the tip will reduce the soil resistance and thus
penetration of the caisson will occur. This required suction reduces the tip resistance and is defined
by Sreq . The relation between the applied suction and the reduced tip resistance is described by
several authors.
The previously discussed methods of Houlsby & Byrne and NGI both assume installation in
homogeneous sandy material. Only Senders [2008] has presented a calculation method for layered
soils. The latter method is appropriate for sand overlaid by clay, however all methods are evaluated

to estimate the reduction in tip resistance due to suction. This is done by implementing a corrected
critical suction for the methods which are used for homogeneous sand conditions (Figure 2.9). In
order to calculate the required suction for installation, the critical suction is taken as upper bound
since the effective stress becomes zero in that case. Calculation methods to determine the reduced soil
resistances are presented in Appendix A.1. The soil resistances without reduction for the previously
discussed methods are presented in Appendix A.3.
Houlsby & Byrne (2005)
According to the simplified method (see Appendix A.3), the tip reduction due to suction is calculated
by taking a factor for the pore pressure into account. The reduction factor for the pore pressures
at the caisson tip for various L/D-ratios takes into account 3D-effects (Houlsby and Byrne [2005]).
For the underlying sand the penetration is given by hs . The dimensionless pore pressure (a1 ) is
calculated as follows;

hs
a1 = c0 − c1 1 − exp − (2.17)
c2 D
Where c0 = 0.45, c1 = 0.36 and c2 = 0.48. As well the effect of inner and outer permeability can
be taken into account, by an adjusted pore pressure factor a(z).
kf a1 ki
a(z) = ; with kf = (2.18)
(1 − a1) + a1kf ko
The final equation is given by equation 2.19, where each component is reduced by a single factor,
varying with penetration depth. When Scrit is filled in as required suction, the remaining variable is
Sred . The equation can then be solved and the minimal required suction can be determined.
W 0 + Sreq Abase = αout (z)Qoutside + αin (z)Qinside + αtip (z)Qtip (2.19)
αout (z) = [1 + a(z)]Sred Ao (2.20)
αin (z) = [1 − a(z)]Sred Ai (2.21)
αtip (z) = f (αout , αin ) (2.22)
Andersen et al. (2008)
The NGI-method for homogeneous dense sands takes into account a ’reduction factor’ for the total
measured resistance by the CPT. This reduction factor depends on the caisson diameter, penetration
depth, soil permeability and critical suction. The normalized critical suction is presented as critical
suction number by Scrit,N GI = Scrit /γ’z, which is calculated according to equations in Appendix A.2.

1
! a(z)
a(z)
0 Sreq
W + Sreq Abase = 1− [Qoutside (z) + Qinside (z) + Qtip (z)] (2.23)
Scrit,N GI
The NGI-method includes a parameter a(z), which varies with the penetration depth. However,
calculating the required suction (Sreq ) is rather complex. Andersen assumes that the reduction-factor
is applied to all soil resistances (inner-, outer- and tip resistance). Based on the Senders’-method
only the inner- and tip resistance will linearly reduce for layered soil conditions. Therefore the
Andersen-method is adjusted by taking a(z)=1 for all z and no reduction of the outer soil resistance
is considered.
Senders (2008)
As assumed by Senders [2008] the inner friction and tip resistance will reduce linearly from the ’initial’
resistance due to push-in installation at zero suction, to zero when suction reaches Scrit,sand . The
critical suction of Senders [2008] is similar to equation 2.16.
!
0 Sreq
W + Sreq Abase = Qoutside 1 − [Qinside + Qtip ] (2.24)
Scrit,sand
2.3.3 Comparison of methods

All methods are compared in Figure 2.10a, where it can be noted that the suction requirement reduces
for the flow condition. The suction requirement is calculated by the critical suction for layered soils.
The CPT-approaches of Andersen et al. [2008] and Senders [2008] both incorporate parameters for
’most probable’ (MP) and ’highest expected’ (HE). These are also presented in Figure 2.10a, where
it can be noted that for the MP-condition the self-weight penetration is approximate 2 m.
0 50
Suction
100
required
150
(kPa)
200 250 300 SFLOW /SNO FLOW
0,0 0,2 0,4 0,6 0,8 1,0
0,00
0,00
Houlsby (No Flow) Houlsby & Byrne
1,00 NGI (MP)
NGI-MP (No Flow) 1,00
NGI (HE)
2,00 NGI-HE (No Flow) 2,00 Senders (MP)
3,00 Houlsby (Flow) Senders (HE)
3,00
Penetration depth(m)
NGI-MP (Flow)
4,00 4,00
NGI-HE (Flow)
5,00 Senders - MP (Flow)
5,00
Senders - HE (Flow)
6,00 6,00
7,00 7,00
8,00 8,00
9,00 9,00
10,00 10,00
(b) Reduction due to flow for different

(a) Required suction for installation without flow (jacked) approaches
and with reduction of soil resistance (flow)
Figure 2.10 Comparison of approaches for suction caisson installation of a typical layered soil profile
(this example presents sand overlaid by 3 m of clay)

For the remaining calculation methods it can be found that the self-weight penetration is lower
(i.e. 1.5 m) due to more conservative input-parameters (higher kp and kf - values). From Figure
2.10b it can be noted that the CPT-based methods show a reduction in the range of 25 - 50%
during the initial penetration in the sand. The Houlsby-method shows a gradual reduction for initial
penetration, but the results approaches the CPT-based results after 2 m penetration. The difference
with the Houlsby-method is likely caused by the limitations of the iterative calculation-procedure and
the adopted simplification of the approach. Within the first meters of penetration the reduction is
higher than for deeper levels of penetration for CPT-based approaches. This is caused by the relative
low critical suction that is applicable for shallow depths.
A remark about the feasibility of both approaches can be made, regarding the limitations of the
chosen calculation methods. The reduced penetration resistances are based on boundary value level
(critical suction), however it is uncertain whether the seepage flow is generated due to plug cracking
or uplift. Further investigation into the mechanisms of plug cracking and plug uplift are further
investigated both experimentally (Chapter 3) and analytically (Chapter 4).
2.3.4 Conclusions
It can be concluded that the approaches for homogeneous sands (Houlsby and Byrne [2005] and
Andersen et al. [2008]) can be adjusted in order to calculate the resistances in layered soils, where
sand is overlaid by clay. The most important adjustment was to correct the critical suction for the
homogeneous sandy material. The critical suction was horizontally translated for shallow penetration
in the sand, according to the solution of Senders [2008].
In the analysis (see Figure 2.10) it was found that during the suction-assisted penetration the total
soil resistance is reduced, but that this reduction gradually decreases with increasing penetration
depth. The reduction at greater depths is roughly 25 - 50% compared with the initial resistance in
the sand. It should be remarked that the reduced soil resistances are based on the assumption of
generated seepage flow, but the exact mechanism should be further assessed (either plug cracking of
plug uplift).
2.4 Conclusions
From practical experiences about installation of suction caissons in layered soils, it was found that
installation is achieved by the use of water-injection-devices. The theory about installation in sand
overlaid by clay identified two mechanisms; cracking- or uplift of the clay plug. From both practical
experiences and theory from literature it can be concluded that fast installation is desired to limit
plug uplift and control the risks involved.
In order to predict the soil resistance during the self-weight penetration phase, limit state meth-
ods are applied. In general there are two options; effective stress- or CPT-approach. The difference
between the methods relies on key-input parameters, such as friction angle and effective unit weight
(effective stress-approach) or cone resistance (CPT-approach).
From an engineering point of view, the CPT-approach is preferred because the relation to the
measured resistance is directly related by a single factor. As well the CPT should encounter the
same soil response (drained, partially drained) as for the caisson during installation . Only the
shape factor is different and should therefore be chosen properly. According to Senders [2008] and
Andersen et al. [2008] it is possible to estimate the reduced stresses during the suction assisted phasel.
The Houlsby & Byrne method is based on effective stresses, which are calculated in an iteratively
way. In addition, this describes the stress changes (reductions) for suction-assisted penetration, which
makes this approach useful. On the other hand the effective stresses are not applicable for layered

soil conditions, thus some adjustments should be made. This is done by introducting a simplified
Houlsby & Byrne method which is less extensive.
Furthermore, it can be concluded that the approaches for homogeneous sands (Houlsby and Byrne
[2005] and Andersen et al. [2008]) can be adjusted in order to calculate the resistances in layered soils,
where sand is overlaid by clay. The most important adjustment was to correct the critical suction
for the homogeneous sandy material. The critical suction was horizontally translated for shallow
penetration in the sand, according to the solution of Senders [2008].
It was found that during the suction-assisted penetration the total soil resistance is reduced, but
that this reduction gradually decreases with increasing penetration depth. The reduction at greater
depths is roughly 25 - 50% compared with the initial resistance in the sand. It should be remarked
that the reduced soil resistances are based on the assumption of generated seepage flow, but the exact
mechanism should be further assessed (either plug cracking or plug uplift). The subsequent Chapter
briefly discusses the plug response during installation (Chapter 3).

Chapter 3
Plug response during installation
This Chapter discusses the literature about plug response during installation of suction caissons
(section 3.1). Since there is limited literature about layered soils (especially sand overlaid by clay)
references about installations in homogeneous sands and clays are considered as well. Based on the
literature review on the plug response during installation, some experiments on plug stability have
been executed at the Geotechnical Laboratory of Delft University of Technology. Information about
the experiments on plug stability can be found in section 3.2. Finally the conclusions of plug response
during installation are presented in section 3.3.
3.1 Theory from literature

Senders [2008] describes that seepage flow can be generated in sand overlaid by clay, by cracking or
plug uplift of the clay layer. The exact failure mechanism is not fully understood, therefore both
mechanisms are considered for further research. In the subsequent analysis plug uplift and plug
cracking is described.
3.1.1 Theory of plug uplift

The potential for plug uplift failure is dependant on the applied differential pressure acting on the plug
in relation to the resistance of the plug. Several side-effects could influence the pressure development,
which will also influence the plug uplift behaviour. For instance, the restricted seepage flow in case
of sand overlaid by clay will be influenced by the surrounding overlying clay properties. In order to
avoid complex correlations, the uplift failure is considered to be primarily dependent on the pressure
acting on the plug. Figure 3.1 presents a schematization of the parameters that determine the plug
uplift criteria.
Di
Suction caisson
pplug
zclay αsu γc αsu Clay
Sand
Figure 3.1 Schematization of plug uplift parameters
21
From the theory of Senders [2008] the plug uplift mechanism is described with equation 3.1. This
equation describes the limiting differential pressure for intact plug uplift, which takes into account
self-weight and skirt friction to overcome, prior to uplift. Hence it can be noted that contribution of
reverse end bearing is not included.

0 4
pplug = γc + αsu zclay (3.1)
Di
Cotter [2009] describes that plug uplift is also dependant on the intact shear resistance of the
’underlying’ clay. In that case the plug will lift up prior to full penetration of the clay layer. The
required uplift pressure depends on the depth of penetration into the clay layer (equation 3.2) and
properties of the clay. In case of full penetration into the clay layer (h = zclay ), the equation becomes
equal to equation 3.1 and intact uplift is assumed.
!
Di2

4 0
pplug = 2 su Di h(α − 1) + zclay γc zclay + (3.2)
Di 4
The driven values to resist plug uplift are the self-weight and inner skirt friction of the clay in
case of full penetration. From equation 3.1 it can be seen that uplift resistance will be higher when
the variables on the right side of the equation increase. Except for the inner diameter (Di ) it holds
that higher uplift resistance will be achieved by a smaller Di . A comparison of both equations is
presented in Figure 3.2. It can be remarked that for relative thick plugs, the ratio D/z is small and
the normalised suction for uplift will be higher.
8,00
7,00
6,00 Senders (2008) plug failure
5,00 Cotter (2009) plug failure for

z=0.5h
P/yc'D
4,00
3,00
2,00
1,00
0,00
1,0 6,0 11,0 16,0 21,0 26,0
D/z
Figure 3.2 Plug failure according to Senders [2008] and Cotter [2009]
3.1.2 Theory of plug cracking

The cracking mechanism is described by Senders [2008] by assuming seepage flow along the skirt of
the suction caisson (piping) or cracking in the middle of the clay. Failure of the piping principle is
described by shear failure at the edges of the clay plug. Piping along the clay skirt will occur when
the applied pressure will be higher than the reduced shear strength. The research of Senders [2008]
does not focus on the failure mechanisms for plug cracking, therefore an estimate of plug cracking
is developed withing the framework of this thesis. A schematization of plug cracking is presented in
Figure 3.3.

Plug response during installation 23
Di
Suction caisson
pplug
Sand
Figure 3.3 Schematization of plug uplift parameters
The phenomenon of cracking can be estimated by calculating the bending moment of the clay
plug. The calculation is simplified by assuming linear elastic behaviour of the clay plug and plate
bending with clamped- or hinged edges. The difference in fixation provides a range of solutions
for cracking failure. The resistance for cracking is calculated using the undrained shear strength as
tensile strength for clay according to Thusyanthan et al. [2007]. Considering the applied cracking
pressure as uniform distributed load, the required cracking pressure range is presented in equation
3.3. A detailed elaboration is presented in Appendix C.
2 2
64su zclay 64su zclay
3 < pcrack < 3 (3.3)
πDi (3 + ν) πDi (1 + ν)
Similar to equation 3.1, it can be seen that increasing self-weight and undrained shear strength lead
to higher pressures for cracking failure. However, higher pressures will be achieved by smaller inner
diameter to the power 3. The relation with inner diameter is much higher in this case, which makes
the driving failure pressure be dependent on the dimensions of the suction caisson. A comparison of
plug uplift and plug cracking is presented in Figure 3.4.
8,00
7,00 Senders (2008) plug failure

Cotter (2009) plug failure for z=0.5h
6,00
Cracking failure, clamped
5,00 Cracking failure, hinged
P/yc'D
4,00
3,00
2,00
1,00
0,00
1,0 6,0 11,0 16,0 21,0 26,0
D/z
Figure 3.4 Plug failure and bandwidth of cracking failure (clamped and hinged)
From both mechanisms it can be found that the dimensions of the plug (i.e. the suction caisson)
determine the driving failure mechanism. Within this framework some series of tests were executed
in the laboratory of Delft University of Technology. The results are presented in section 3.2.

3.1.3 Reference data of plug response

Considering the two mechanisms described, it can be remarked that the dimensions depend whether
the plug will lift or crack during the suction process. According to several references (see Figure 3.5)
it has been found that plug uplift did occur in cases for relative thick plugs. From a practical point of
view the range of reference data is limited to D/z < 6, since thinner clay layers are not considered in
practice. In order to analyse the data, an upper- and lower boundary of the undrained shear strength
is chosen; i.e. lower boundary of 5 kPa and upper boundary of 50 kPa.
8,0 Plug failure| Su=5 kPa

Plug failure| Su=50 kPa
7,0
45 Aas & Saue (2009)
6,0 Allersma (2001)
Cotter (2009)
5,0
Tjelta (1986)
P/yc' D
4,0 Tran et al. (2007)

25 Watson et al. (2006)
3,0
1.7
5.5 5.5
2,0
1,0
0,0
0,0 5,0 10,0 15,0 20,0 25,0 30,0
D/z
Figure 3.5 Plug failure for several reference-data (data-labels represent su [kPa])
Figure 3.5 shows a reasonable bandwidth about reference-data from literature. Some comments
about the data-point can be made:
• The test performed by Allersma et al. [2001] showed that plug uplift occurred for a Kaolin clay
with su = 1.7 kPa.
• Cotter [2009] performed several tests on layered soil strata which consist of sand overlaid by
clay. The shear strength was briefly determined (5.5 kPa) and model-dimensions are presented
as well. The left orange data-point didn’t show any plug uplift, which corresponds to plug
failure boundary.
• The blue data point (Tran et al. [2007]) is well reported, but these tests were performed with silt
materials. Uplift failure did occur, but since silt material is more favourable for scouring-effects
compared with clay, this data point is less guiding.
• The grey data point (Watson et al. [2006]) presents the data reported in Senders [2008], which
briefly described the plug uplift mechanism. However, the undrained shear strength of the
Kaolin-clay used for the centrifuge modelling test was not presented. If a su of 2 - 4 kPa is
assumed (Burns et al. [2010]), the data point is in accordance to the given boundaries.
From all reference data considered, it can be concluded that plug uplift failure is in accordance
with the theory of Senders [2008]. On the other hand, there’s no reference data present about possible
cracking failure, since D/z > 6 is not found in practice. In the framework of this, better understanding
about the failure mechanisms involved was desired. Therefore some small scale tests were set-up in
the laboratory of Delft University of Technology, whether to check this hypothesis. A description
about the test is presented in section 3.2.

3.1.4 Conclusions
From literature it can be found that plug uplift is the mayor failure mechanism to be expected in
all cases. Failure mechanism of cracking has been described by Senders [2008], but has not been
investigated for varying dimensions of suction caissons. By considering clamped and hinged edges of
the plug-disk, a range of possible cracking failure mechanisms can be found. However, verification of
this theory is desired which can be done by some basic experiments on plug cracking. The purpose
of the tests is to investigate the failure mechanisms that occur for different dimensions. As a result
of this, it is possible to determine the transition point or range of failure behaviour, which might
be interesting for the industry. More details about the experimental analysis are presented in the
subsequent section.
3.2 Experiments on plug response

As described in the previous chapters some basic tests are recommended to check the hypothesis
about cracking failure for relative thin plugs. First the possibility of scaled modelling is investigated,
followed by a test set-up. In total 4 series of tests were conducted, whereas each series consists of
two tests.
3.2.1 Scaling laws

Considering available time and budget it was chosen to perform basic 1G-experiments in the labor-
atory of Delft University of Technology. The purpose of the tests is to verify the predictions of
the model and to check whether cracking or plug uplift will occur. Both mechanisms could behave
differently in scaled circumstances; therefore the scaling factors have to be implemented (Table 3.1).
Table 3.1 Scaling laws for 1G testing
Parameter Scaling factor (model/prototype)

Linear dimension (zclay and Di ) 1/n
Undrained shear strength ( su ) 1
Submerged unit weight of clay (γc0 ) 1
Suction pressure (p) 1
Adding these scaling factors into the equations found in previous section, results in;
4
pplug = γc0 nzclay + αsu nzclay (3.4)
nDi
64su n2 zclay
2
64su n2 zclay
2
< pcrack < (3.5)
πn3 Di3 (3 + ν) πn3 Di3 (1 + ν)
It can be seen that both mechanisms are not scaled by pressure or undrained shear strength
(equation 3.4 and 3.5). This implies that the scaling is dependent on the dimensions of the caisson
(or plug) and therefore a fundamental difference in failure mechanism can be investigated, i.e. plug
uplift or cracking. The difference in failure mechanism will be described by a dimension analysis,
depending on the mobilized shear strength and suction pressure. As a result of this, it might be
concluded that scaled model tests can be executed at 1G for these mechanisms in clay. One should
note that for the underlying sand, the calculations of effective stresses don’t scale with the same rates
and thus are not considered.

Parameter determination
Allersma et al. [2001] tested a layered soil profile, consisting of sand overlying by clay and performed
installation of a suction caisson using an oscillating pressure inside the caisson. In order to verify this
theory, similar material properties were taken into account. Taking the prototype-data of Allersma
et al. [2001] with the same Kaolin-clay, the desired undrained shear strength is 10 kPa. A distinction
of uplift or cracking failure was made by taking D/z-ratios far off the transition range 6 < D/z <
10 (see Figure 3.4). Plug uplift failure was tested at lower ratios (D/z ≈ 6) and cracking failure
at higher ratios (D/z > 20). The latter ratio (cracking) is an extreme value from a practical point
of view and will be discussed later. The test apparatus has an inner diameter (Di ) = 190 mm and
therefore determines the D/z-ratio for crack or uplift failure (Table 3.2).
Table 3.2 Model properties
Allersma et al. [2001] Model tests

Prototype Ng scale (50 g) 1g scale 1g scale
(plug uplift) (cracking)
Diameter (Di ) 3m 65 mm 190 mm 190 mm
su [kPa] 23 (K122) 23 (K122) ≈ 6 (Kaolin) ≈ 6 (Kaolin)
10 (K147) 10 (K147)
1.7 (Kaolin) 1.7 (Kaolin)
Thickness (zclay ) 1.7 m* 34 mm > 30 mm < 10 mm
Ratio D/z [-] 1.76* 1.91 mm < 6.3 > 20
* = Back-calculated
Using sensitivity analyses for su , γc0 , α and ν, the upper and lower boundaries were determined
for layer thickness of the clay. The values of layer thickness in Table 3.2 are used as boundaries for
the experiments.
Clay preparation
From Table 3.2 it can be remarked that the undrained shear strength of clay is 6.1 kPa, which is rather
lower than the desired 10 kPa. Assuming sufficient time to obtain U > 99 % as degree of consolidation,
the shear strength of the clay only depends on the preloading stress prior to consolidation. The
limiting factor on the shear strength was given by the weight capacity of the consolidation apparatus
(limited to 75 kg max). Therefore the maximum shear strength to be created was limited. The
actual undrained shear strength was determined by the Vane Test according to BS 1377:part 7:1990.
A series of 3 tests (within a range of 5%) showed that the average peak strength of the clay is 6.1 kPa.
Since for all tests the same conditions were adopted (preloading stress and sufficient consolidation
time), it can be assumed that the undrained shear strength is the same for each test.
3.2.2 Test set-up

The purpose of the tests is to identify the possible transition point or range of the cracking failure for
high D/z- ratios. This is done by taking D/z-ratios far off the predicted transition point. First a series
of observational tests is executed to verify the zone of D/z for plug uplift or cracking. Thereafter the
testing-procedure is optimized and data has been collected for further analysis.
Execution of tests
In the first phase the Kaolin-clay slurry was consolidated by two-way drainage and sufficient pre-
consolidation surcharge. Two-way drainage is established by opening the valve to the water column
(see Figure 3.6). Prior to testing, the water on top of the clay is extracted to perform observational
tests. The level of the water column is kept equal to the level of the saturated clay layer.

Each transparent box filled with sand/clay material is connected to a water column of 1 m height,
which is filled with deoxidized water. As well this water column can be lowered by opening a valve, to
equal both water levels within the soil-box and water column. The principle for testing the stability
of the clay plug for uplift or cracking failure consist of increasing the hydraulic pressure below the
clay layer, by increasing the height of the water column.
Optionally, the valve in between soil-box and water column can be closed, to ensure a larger dif-
ferential pressure that initiates more sudden. Once the clay-slurry was consolidated, the preloading
was removed and increasing differential pressure made the clay samples crack or uplift.
Di,wc
Di
hwc p0
zclay
hbox
p0
Soil box Valve Water column

(a) Test apparatus (b) Schematization of set-up
Figure 3.6 Set-up of tests
A schematization of the test-apparatus is presented in Figure 3.6b, including dimensions. In Figure

3.6a the schematization of the the soil-box is shown, which consist of a permeable material overlaid
by the consolidated clay with height zclay . The hydraulic gradient wihtin the soil-box is concentrated
in the centre of the testing cylinder. The porous filling material will spread this concentrated flow,
ensuring a uniform distributed pressure below the clay layer.

Pressure distributions
Pressurization of the clay layer is achieved by increasing the hydraulic head acting below the clay,
which is done by filling the water column. The difference in water table of the soil-box and water
column is defined by the water table difference (∆hwt ). Depending on the magnitude of ∆hwt , there
can be high or low pressures subjected to the clay sample. Depending on the conditions of a large or
small ∆hwt , the pressures on the clay are calculated differently;
Large difference condition
A large difference in water table causes a high ’potential’ pressure on the clay sample. However,
the water inflow into the soil-box is limited by the dimensions of the connecting tube. This means
that the pressurization of the clay is limited by the inflow-rate of the tube. Thus the pressure-rate
on the clay is not given by the water column, but depends on the maximum inflow-rate of the tube
(for large ∆hwc ). In order to apply a large ∆hwc on the clay sample, the valve is closed while the
water table of the water column is filled.
Small difference condition
In case of a small difference between the water tables, the differential pressure on the clay is low.
There is more time required to level both water tables, but the flow-restriction of the connecting
tube is not applicable anymore. The pressure on the clay is now given by the actual height of the
water column. In order to make a distinction between the two conditions, the large difference holds
for the case when ∆hwt > 30 centimetres. Below this value is the condition considered to be a small
difference (see Figure 3.7).
70
Large
60
difference
50 Boundary
Δwt [cm]
40 test 2a
30 test 3
test 4a
20
Small test 4b
10
difference
0
0 50 100 150 200
Time [s]
Figure 3.7 System-permeability test without clay

Literature Review and
Literature Experiments
Review and Experiments Remon Romp
Remon Romp
Literature Reviewand
Literature and Experiments
Review and RemonRomp
Romp
Literature
LiteratureReview
Literature
Review and
Review and Experiments
Experiments
Experiments
Experiments Remon Remon
Remon Romp
Remon Romp Romp
3.2.3 Results
4.3 Results
4.3 Results
4.3 Results
4.3
4.3 Results
This4.3
4.3 Results
Results
section
This section
Results
presents the results of theoftests performed at the Delft University of Technology labor-
This This presents
section presents
the results
section presents
the the of
results
the tests
results
the tests
performed
of the
performed
at the Delft
tests performed
at the Delft
University
at the
University
of
Delft University
of of
atoryThis
This This
section
section
This
Technology section
presents
of Geo-Engineering.
presents presents
the
of results
The
sectionlaboratory
laboratorypresents
the the
first
results results
ofof
test
the the of
tests
series
results
the
Geo-Engineering. the tests
tests performed
performed
ofrepresent
teststhe
performedthe
performed at
at the
atatobservational
the
theDelft
Delftthe Delft
tests.University
University
University
Delft of
In of
of the subsequent
University of
Technology
Technology laboratory of Geo-Engineering.
Technology
analysis Technology
Technology laboratory
Technology
laboratory
the results ofof Geo-Engineering.
laboratory
laboratory
of series of Geo-Engineering.
of2Geo-Engineering.
Geo-Engineering.
- 4 areof presented.
Geo-Engineering.
4.3.1 Test4.3.11 – Test
Observational
1 – Observationaltest test
4.3.1 Test
Test11
4.3.1 – Test
Observational
1 test
4.3.1
4.3.1 Test4.3.1 Test 1 –– Observational
1––Observational
Observational test
Observational
test test
test
For
Test For both thin
For
1 Observational and
both thick
thin and
test clay plugs
thick clay the p\ressure
plugs of the water
the p\ressure of thecolumn was increased
water column was increased
both thin and thick clay plugs the p\ressure ofof thewater
water column wasincreased
increased
For bothFor
Forboth
stepwise. thin
For
thin
In both
and
both
and
both
stepwise.
thin
thick
thin
thick
test
In
and
clay
and
series
both
thick
clay plugs
thick
plugs
test
clay
the the
clay the
rate
series
plugs
of
the
the
p\ressure
plugs
p\ressure
the p\ressure
the
p\ressure
pressure
rate of of the of
water
increase
pressure of the
the water
column
column
waswater
increaselow. column
was
column
was
was thewas
increased
Forlow. was
thick
For
increased
increased
the thick
stepwise. Inboth
bothtest
stepwise. test
In series
both test the ratethe
series ofpressure
pressure
rate of increaseincrease
pressure waslow.
low.was
Forlow.
thethick
thick
For the
stepwise.
stepwise.
clay plug
For both thinclayIn
stepwise.
In both
(Figure
andplug In
test
9, series
both
series
left)
thick(Figure
clay plugsittest
9,
the
was the rate
series
rate
clearly of
the
of
the itpressure
left) was pressure
rate of
observed increase
pressure
increase
that
of theobserved
clearly plug
water column was
increase
thatwas
upliftlow.For
wasFor
occurs,
was increased
plug uplift
the
low.
the For
whilethick
the the thick
stepwise.
occurs, while thick
In
theboth
clayplug
clay plug (Figure
clay plug
(Figure 9,9, left)itit9,
(Figure
left) was
was left) clearly
it
clearlywas observed
clearly
observed thatplug
observed
that plug thatuplift
plug
uplift occurs,
uplift
occurs, whilethe
occurs,
while thewhile the
clay
test series plug
theclay
(Figure
stays
clay
rate plug
plug
of 9,
(Figure
intact left)
stays during
pressure it
9,
intact was
left) clearly
it
uplift.
during
increase was
During
was observed
clearly
uplift.
low. observed
uplift
During
For that
there
the plug
was
uplift
thick thatuplift
no
there
clay plug occurs,
uplift
leakage
was
plug no occurs,
while
observed
leakage
(Figure 3.8 the
onwhile
observed
- left the
on it
series)
clayplug
clay plug stays
clay
stays plugintact
stays
intact during
intact
during uplift.
during
uplift. During
uplift.
During uplift
During
uplift there
there was
uplift
was noleakage
there
no leakage
was no observedobserved
leakage
observed onon on
clayofplug
top clay
the stays
topclay
ofplugintact
stays
plug.
the clay during
intact
plug. uplift.
during During
uplift. uplift
During there
uplift
was there
no leakage
was no leakage
observed observed
on on
was top
clearly observed
topofofthe
the
topclay
of that
plug.
the clay plug uplift occurs, while the clay plug stays intact during uplift. During
top of the clay
top ofplug.
clay the clay plug.
plug. plug.
uplift there was noTest 1a Test
leakage |D/z
observed = 6.3 on =top 6.3of the clay plug. Test 1b|D/z = 19
Test1a |D/z1a
1a Test ==
|D/z
6.3 Test1b|D/z Test 1b|D/z
1b|D/z =19
19 = 19
Test
Test 1a|D/z |D/z1a
Test 1a |D/z
|D/z =
=6.3
6.3 = 6.3
6.3 Test
Test 1b|D/z Test
Test =1b|D/z
1b|D/z
= 19 =
= 19
19
Initial phase
Initial phase Initial phase
Initial phase
Initialphase
phase
Initial Initialphase
phase
Initial Initial phase
Initial phase phase Initial Initial
Initial phase
Initial phase phase
(a) Initial phase (b) Initial phase
Uplift of 2Uplift
mm of 2 mm Uplift of 0,5 mm
Uplift of 0,5 mm
Upliftofof22Uplift
mm of 2 mm Upliftofof0,5
0,5 mm
Uplift of
Uplift mm
Uplift of 2Uplift
mm of 2 mm Uplift
Uplift of 0,5 mm
Uplift
mm of 0,5 mm
0,5 mm
(c) Uplift of 2 mm (d) Uplift of 0.5 mm
Uplift of 1Uplift
mm of 1 mm
Uplift of 6Uplift
mm of 6 mm Upliftofof11Uplift
mm of 1 mm
Upliftofof66Uplift
mm Uplift
Uplift mm
(e) Uplift
Uplift of
Uplift mm of
mm
of 66Uplift 6 mm
of 6 mm (f ) Upliftmm
of 1Uplift
of 1ofmm
1 mm
Uplift of 2Uplift
mm of 2 mm
Uplift of 12mm Upliftofof22Uplift
Uplift mm of 2 mm
mm
Uplift of 12mm Uplift
(h) of 2Uplift
Upliftmm of of 2 mm
2 mm
Upliftof
(g) Uplift ofof12
12mmmm 12mm
Uplift
Uplift Uplift of
Uplift of12mm
12mm of 12mm
Figure 9: Figure
Uplift of9: clay forofrelative
Uplift clay forthick (left)
relative and (left)
thick thin (right)
and thinclay
(right) clay
Figure 9:Uplift
Uplift
ofofof clay forofrelative thick (left) andthin
thin (right) clay
FigureFigure
Figure 9: Figure
3.8 9:Uplift
Figure
Uplift 9:
of
9: Uplift
clay
clay
clayfor
for
Uplift
forof clay
clay for
relative
relative
relative
for relative
thick
thick
thick(left)
(left
relative thick
and
(left) and (left)
series)
thick and
(left)
thin and
(right)thin
clay
thin
(right)
and thin (right)
(right clay
(right)series)
clay clay clay

Delft University of Technology 16 16 SBM Offshore
SBM Offshore
Delft
Delft
In University
Delft
theUniversity
case ofTechnology
of
thinTechnology
of aUniversity
clayof Technology 16
(Figure 3.816 16 SBM
uplift of theSBM Offshore
SBM
clay Offshore Offshore
Delft University
Delft University oflayer
of Technology
Technology - right series)
16 16 SBM layer was
Offshore
SBM observed
Offshore
as well due to gradual increasing pressure. However, after further increase of the water table in the
water column (high ∆hwt ), the plug starts to bend with approximate 5 mm deflection in the centre of
the plug. The convex-shaped plug moved further upwards until cracking occurred due to shear forces
(see Figure 3.9). The cracking causes a minor leakage, since it didn’t cause an increased pressure
drop of the water column.
The first test-series showed plug uplift for both high and low D/z-ratios, but only for gradual
increasing pressure. The locations of the cracks indicate local shear failure of the clay layer due to

centre).
The convex-shaped plug moved further upwards until cracking occurred due to
shear forces (see Figure 10). The cracking causes a minor leakage, since it didn't
cause an increased pressure drop of the water column.
Figure 3.9 TopFigure 10: Top view of cracks with minor leakage
view of cracks with minor leakage (≈ 1 % of cracks)
The first test-series showed plug uplift for both high and low D/z-ratio, but only for
gradual increasing pressure. The locations of the cracks indicate local shear failure
the mobilizedofpressure. Despite
the clay layer duethe cracking
to the observed,
mobilized still Despite
pressure. the initial
the failure mechanism
cracking observed,for
stillboth
the was
clay thicknesses initial failure mechanism for both clay thicknesses was uplift.
uplift.
4.3.2 Test 2
Test 2 In the second test series, a plastic membrane was used in order to create a high
permeable base. By the consolidation of the clay-slurry, it showed that the plastic
In the second test series, a plastic membrane was used in order to create a high permeable base. By
membrane was compressed more than 50%. Once the top-load was removed, the
the consolidation of the
elastic clay-slurry, itofshowed
de-compression that the
the plastic plastic membrane
membrane made the was claycompressed
lift op. Once morethethan
50%. Once the membrane was in equilibrium again, small pressure increments lead to uplift of the the
top-load was removed, the elastic de-compression of the plastic membrane made
clay liftLiterature
clay Review
up. Once plug, and Experiments
possibly
the membrane by
wastheinmobilized
equilibriumremoulded shear
again, small strength
pressure Remon
along
increments theRomp
skirt.
lead to In
uplift
the end
of the clay plug, no reliable
possibly by thedata could be
mobilized implemented
remoulded shearbystrength
this test.
along the skirt. In the end no
reliable4.3.3 Test 3be
data Despite
could and 4
implemented by this test.
there was no reliable data obtained, still some observations were made
Despite there
during
This test was no reliable
testing.
series For of
consist data
a gradual
a moreobtained,
solid,still
increasingbutsomestill observations
water werethat
table, itpermeably
highly showed made theduring testing.
clay plug
underlying
For a gradual lifted
material upwards.
(gravel),
increasing While
overlaid
water bythe
table, water
clay
it withtable
showed of the
thethat
same water
clay column
theproperties
plugas wasupwards.
furtherbefore.
mentioned
lifted increased,
WhileIn thethe
water
table oforder clay
the water plug
to avoidcolumn cracked
the whenincreased,
clay further
was penetratingΔwt into
wasthe
thelarge.
gravel
clay This shows
during
plug cracked that ∆hthe wtrate-effect
consolidation,
when awas large. ofThis
geo-
shows that pressurization
membrane
the is used of
rate-effect toalso depends
separate the which
pressurization clay failure
alsofrom the mechanism
depends gravel.
which For could
this
failure beseries
test expected.
mechanism the valve
could be expected.
to the soil-box was closed and the water column was filled till Δwt ≈ 70 cm (large
difference condition).
Test 3 and 4
During the second test (test 4a), it turned out that the geo-membrane contributes to
the series
These test entireconsist
uplift of
of the claysolid,
a more layer. Simultaneously,
but there leakage
still highly permeably started
underlying along(gravel),
material the over-
skirts,Delft
due to uneven uplift
University of of the plug.
Technology 17 SBM Offshore
laid by clay with the same properties as mentioned before. In order to avoid the clay penetrating into
In the
the gravel firstconsolidation,
during and third test the clay didn'tislift
a geo-membrane up,
used to but startsthe
separate leaking once
clay from the
the high For these
gravel.
differential
test series the valvepressure was initiated.
to the soil-box was closedIn and
all test the clay
the water cracked
column wasjust
filledbefore the
till ∆h wt water
≈ 70 cm (large
column
difference dropped
condition). for 5 centimetres.
During the second testPrior to 4a),
(test the cracking
it turnedprocess
out thatbulging of the clay contrib-
the geo-membrane
utes to was observed in test 3 and 4b.
entire uplift of the clay layer. Simultaneously, leakage started along the skirts, due to uneven
uplift ofFor
theallplug.
tests it was observed that leakage starts at the skirts of the clay, followed by
some cracks in the middle of the clay as indicated in Figure 11.
Figure 11: Top view of cracks for test 3 (left) and test 4b (right)
Figure 3.10 Top view of cracks for test 3 (left) and 4 (right)

In the first and third test the clay didn’t lift up, but starts leaking once the high differential
pressure was initiated. In all test the clay cracked just before the water column dropped for 5
centimetres. Prior to the cracking process bulging of the clay was observed in test 3 and 4b. For
all tests it was observed that leakage starts at the skirts of the clay, followed by some cracks in the
middle of the clay as indicated in Figure 3.10. This can be justified by the reduced adhesion of the
clay-perspex interface and the maximum bending moments, which indicates the weakest points of
the clay plug.
Back-calculations
In case of the third test series the water table of the water column and time are accurately measured.
The data from the measurements (time and water table heights) are used to back-calculate the pres-
sure on the clay during failure. To do this accurately, a lower-/upper boundary approach has been
chosen, which is presented below;
• Lower boundary: System permeability
Prior to the tests on the clay layers, the time required to fill the soil-box by opening the valve of
the water column was measured (Empty in Figure 3.11). This could be regarded as the ’system-
permeability’. This system-permeability is used as a reference, to compare with the system permeab-
ility during testing (With clay in Figure 3.11). For both situations (reference and testing) it holds
that the water table of the water column was set to 0.90 cm, prior to opening of the valve.
18,3
18,1
H_box [cm]
17,9
Empty
17,7
17,5 With clay
17,3 5 cm drop
17,1 Video
16,9
0 20 40 60 80 100
Time [s]
Figure 3.11 Reference permeability for test 3
In case of the soil-box being empty (water only), the increase of the water table in time is
measured. As mentioned in previous section, the inflow-rate depends on the dimensions of the tube
(large difference condition). Without any resistance, the increase of water table is given by Empty
in Figure 3.11. During testing, the increase of water table is ’blocked’ by the clay layer, which
causes more resistance. As a result the pressure drop in the water column is ’delayed’. However, the
maximum pressure-rate stays similar to the empty situation, so the equivalent pressure of cracking
can found by taking the reference ∆hbox at time of failure. The value of ∆hbox at failure can then
be back-calculated to the actual pressure at failure. The time required to cracking failure is recorded
(Video in Figure 3.11) and an intermediate data point has been added when cracking started.

• Upper boundary: Volume continuity
The pressure increment on the clay can also be calculated by taking volume continuity into
account. All tests have been executed with deoxidized water and air capture was avoided, which
contributes to straight-forward continuity calculations. Therefore the extracted volume of water in
the water column (Vwc ) will be equal to the increased volume of water in the soil-box (Vbox ). The
effect of evaporation is assumed to be disregarded. While the water table in the water column is
lowered, the equivalent displaced volume needs to be transferred to the soil-box, which causes a
pressure increment below the clay layer. In case of a 5 cm drop an increment of 0.07 kPa in the
soil-box is to be expected (0.68 cm). Since all tests failed (i.e. uplifted or cracked) just before the
first measurement-point at 0.85 cm, this estimate will be the upper boundary of the test results.
Table 3.3 Bandwith of pressures below clay
Test No. Lower boundary [kPa] Upper boundary [kPa] Range [kPa]
3 0.055 0.068 0.013
4a 0.095 0.068 0.027
4b 0.120 0.068 0.052
As can be found in Table 3.3, the ranges of the lower and upper boundaries are very small. For this
reason only the range of test 4 is taken into account for this approach. From Figure 3.12 it can be
found that the test-data are in accordance with the cracking predictions. It should be noted that
the horizontal axis is bounded by D/z > 20. Further discussion about the results is presented in the
next section.
1,0
Test 1b - Plug + Crack
0,9 Test 3 - Crack
Test 4a - Uplift
0,8
Test 4b - Crack
plug
0,7 Plug failure| Su = 6.1 kPa
0,6
Cracking failure, hinged
P/yc' D
0,5
0,4
0,3
crack
0,2
0,1
0,0
20,0 25,0 30,0 35,0 D/z 40,0 45,0 50,0 55,0
Figure 3.12 Test results of high D/z-ratio compared with model (including bandwith)
3.2.4 Conclusions & discussion

From estimations of video-recordings, a back-calculation of the observational tests (test 1) can be
madel. From the 3 tests that were implemented for relative high D/z-ratios, the measurements are
back-calculated to compare with the model. All test results are presented in Figure 3.13 with a full
range of D/z-ratio.
It can be concluded that the failure mechanism is indeed dependant on the dimensions, since a
clear distinction in required pressures was measured. For relative thin clay plugs the tendency of
failure is cracking, while for thicker plugs uplift occurred. The first test series (test 1) encountered
plug uplift for both D/z-ratios, which fit to the plug failure prediction in Figure 3.13. In case of test

8,0
Test 1a - Plug
7,0
Test 1b - Plug + Crack
6,0 Test 3 - Crack
Test 4a - Uplift
5,0
Test 4b - Crack
P/yc' D
4,0 Plug failure| Su = 6.1 kPa
3,0
Cracking failure, hinged
2,0
1,0 plug
0,0
crack
0,0 10,0 20,0 30,0 40,0 50,0 60,0
D/z
Figure 3.13 Test-results compared with model, including diameter (D) and thickness of clay plug (z)
1a and 1b-plug, the differential pressure was low (small difference condition), which made the plug
lift upwards in both cases. In test 1b-crack, 3 and 4 the large difference condition was applied, which
resulted in cracking failure in all cases. As the transition range is 6 < D/z < 10, it showed that for
relative high D/z-ratios cracking was the governing failure mechanism, which is in accordance with
the predictions.
However, in case of test 1b it can be seen that for a high D/z-ratio (test 1b) still uplift occurred
which is close to the prediction of plug uplift failure. This data was obtained while the pressure-rate
on the clay was low (small difference condition). But when the condition on the same plug was
adjusted to a large difference condition, the failure mechanism changed as cracking was observed.
This change of failure mechanism was also observed for test 2, which did not provide reliable data
due to elastic deformations of the underlying material during consolidation.
From this it can be concluded that the rate-effect of pressurization also depends on which failure
mechanism could be expected. Even for high D/z-ratios it was found that uplift occurred, but only
in case of small pressure differences.
Idealized boundary conditions
In general it can be remarked that the calculation model is strongly idealised, which also includes
idealised boundary conditions (i.e. horizontal layering, uniform shear strength of clay, uniform thick-
ness of clay layer and fully saturated clay and underlying material). In case of plug cracking, the
assumption about clamped or hinged edges was made to provide a range of solutions. The calculation
is based on static mechanical loading on a circular plate. Because the clay is a cohesive material, the
assumption of plate bending could be made. From observations it was found that the bulging indeed
developed and therefore this schematization was in accordance with the observations. However, the
behaviour of the clay material is rather complex in case of failure at large deformations. Some plastic
deformations are expected, which are not taken into account within the calculations.
Pressure calculations
The calculations to indicate the pressure below the clay layer are based upon static back-calculations
of the pressures. As a measure of back-calculations, the pressures are captured by using a lower-
/upper boundary approach. More detailed experiments are required with monitoring equipment to
measure the failure pressures more accurately.

Pressure instead of suction
In order to obtain a differential pressure on the clay plug, the hydraulic pressure below the clay plug
is increased, while the atmospheric pressure acts on the top of the clay layer. In both situations
the differential pressure makes the clay move upwards. Pressurization of the clay is applied during
testing, while in practice suction acts on the clay. This adjustment excludes generation of negative
excess pore pressures below the clay plug (principle of reverse end bearing), which is therefore not
considered. As well stress-dependency of clay in (deep) sea conditions are expected in practice. The
limitations of these aspects on the experimental results should be therefore be recognized.
Time dependency
From the test results it has been found that the failure mechanism also depends on the pressure-rate.
Senders [2008]) also described a rate-dependency for potential plug uplift by equation 3.6.
!
p2 pplug
Lplug,lif t = ksand − kplug ∆t f or p2 ≤ p1 − pplug (3.6)
γw save γw Hplug
For ksand >> kclay and large ∆t is shows that the plug heave (Lplug,lif t ) increases while the
pressures are kept constant. This rate-dependency was also observed during centrifuge experiments
by Watson et al. [2006]. For these tests the pressure-rate has a similar effect on the clay layer, as the
suction-rate has during installation in practice. From this point of view, it is interesting to further
investigate to rate-dependent behaviour of plug stability (see Chapter 4).
Practical perspective
From a practical perspective it can be remarked that the encountered range of interest when clay tends
to crack (D/z ≈ 15) is rather thin. In case of an inner diameter of 10 m, the tendency of cracking will
occur in layers with thickness lower than 0.7 m. Plug uplift would therefore be expected for industry
practice values. However, the limitations of the theory shall be considered (e.g. contribution of
reverse end bearing). From the data which has been collected, both the quality and quantity are
not sufficient to present adjustments for the current design method. However, still some conclusions
could be drawn, regarding rate- and dimension dependency. In this framework, it is proposed to
further investigate the rate-dependency of the plug uplift behaviour (see Chapter 4).
3.3 Conclusions
According to the literature it has been found that in general two types of failure mechanisms are
involved for installation of suction caissons in sand overlying by clay; plug uplift or plug cracking.
Comparing these two failure mechanisms it showed that the failure mechanism depends on the dimen-
sion of both the caisson (D) and clay thickness (zclay ), clay properties (su and γc0 ) and rate-effects.
The latter is observed from several experiments (Senders [2008] and Watson et al. [2006]), however
not well documented in the guidelines.
From literature research it can be concluded that cracking of the clay plug is a non-existing fail-
ure mechanism in practice nowadays. Considering typical cases with D/z ≈ 6, it was assumed that
plug uplift occurred during installation, however there’s an uncertainty at the transition point re-
garding failure mechanism, i.e. cracking or uplift. This indicates a limitation of the theory and is
further investigated within the framework of this research.

In order to implement these two mechanisms into a model to predict plug uplift or cracking fail-
ure, the required failure pressures were calculated for different D/z-ratios. The uplift calculation
of Senders [2008] was used to predict the uplift failure. With the use of schematization of circular
plate bending, an estimation of the cracking failure could be made. From the model a transition of
plug uplift or cracking was noticed for 6 < D/z < 10, where uplift was expected for lower D/z-ratios.
However, from a practical perspective it can be remarked that the encountered range of interest when
clay tends to crack is rather thin.
Within this framework some basic experiments were conducted at the Delft University of Tech-
nology to investigate the failure mechanisms for varying D/z-ratios. The D/z-ratios of the tested clay
samples were far off the transition range of plug or cracking failure, to ensure a distinction in failure
mechanism. After back-calculation of the test results, it can be concluded that the model predictions
are in range of the failure pressures. Cracking failure did occur for high D/z-ratios, which actually is
in accordance with the theory. However, from observations of the same clay plug with high D/z-ratio
there also was plug uplift encountered. This occurred for a low differential pressure condition, which
indicates a pressure-rate dependency of the failure mechanism. This can also be found in literature
(Senders [2008] and Watson et al. [2006]), where slow penetration (low differential pressure) contrib-
utes highly to uplift of the clay. In addition to this it was observed that high differential pressures
contribute to cracking of the clay layer during testing. The relation of installation-rate and uplift of
the clay plug is further discussed in Chapter 4.
The test results on plug stability exclude the possible contribution of reverse end bearing in the
underlying sand, prior to uplift or cracking of the plug. The theory of reverse end bearing in layered
soils is presented into detail in Chapter 5.


Chapter 4
Assessment of plug stability
In the considered framework an assessment of the plug stability is realised by introducing two extreme
cases; penetration in layered soil with a stable plug (Figure 4.1a) or a moving plug (Figure 4.1b). It is
assumed that ’some’ seepage is required to install the suction caisson to target depth. The red-arrows
indicate the considered reduced soil resistance due to the seepage flow. The first case is described
in section 4.1, the second case is discussed in section 4.2. Finally the conclusions of both cases are
presented in section 4.3.
ΔS ΔS1
Plug
Clay ΔS2 Clay
Sand Sand
(a) Seepage through stable clay plug with cracks (b) Uplift of intact clay plug
Figure 4.1 stable and moving plug
Case 1: Stable plug
If it is assumed that reduction of tip resistance is required, the seepage through the clay should be
sufficient to induce a seepage flow in the underlying sand. For an intact plug this means that the
permeability has to increase. The permeability of the clay plug can increase due to cracks in the
clay plug. The critical seepage is described by a ’critical permeability’, which can be related to a
fracture-area of the clay.
Case 2: Moving plug
In case of a moving plug, the seepage flow is induced by the uplift of the clay plug. The suction
below the clay plug will attract water (generating seepage) and thus reduction of the tip resistance
should occur.
37
4.1 Installation with a stable plug

When uplift of the plug should be prevented, the (boundary) condition of a stable plug is considered.
From the theory obtained in Chapter 2 it was found that cracking of the plug is not considered
as failure mechanism during installation. According to basic experiments on plug response it was
back-calculated that a small amount of cracks (≈ 1 %) resulted in substantial seepage flow. The
principle discussed in this section is to assess plug instability by increasing the plug permeability.
This increased permeability (critical permeability) can then be rewritten by introducing a fracture-
area; the clay plug will be fractured and the plug permeability has to increase. Because of the upward
seepage flow it is assumed that the cracks are filled with the underlying sand material. As a result
the permeability of fractured clay plug will increase, because of the more permeable sand (compared
to clay).
4.1.1 Seepage through clay plug

The seepage through the clay plug is similar to a 1D-permeability problem. There will be relative
negative pore pressures due to the applied suction. With regards to the tip reduction due to seepage
flow, the question arises to what extent the suction applied above the clay plug will also result in suc-
tion just below the plug (Figure 4.2). For impermeable clay layers with permeability of k < 10−5 m/s
it was found that there is negligible pore pressure reduction below the clay plug within the typical
time span of caisson installations (see Appendix D.1).
S P (kPa) P (kPa)
Suction Suction
caisson S
caisson
Clay Sred Clay
d(m) d(m) αSred

Sand
Sand
(a) Pressure drop over the clay layer (b) Pressure drop through clay and sand
Figure 4.2 Suction inside caisson with idealized pressure drop
Critical permeability
In order to estimate the selected suction for fractured clay, an approach considering the critical
permeability is chosen. This critical permeability is based on the permeability of the intact clay, but
will be higher in case of cracks. It is assumed that the cracks are filled with the underlying sand
material, so the critical permeability can be calculated with;
S
Qcrit = Qclay + Qsand = kclay Aclay + ksand Asand (4.1)
hc γw
S
Qcrit = kcrit Abase (4.2)
hc γw

Plug stability 39
kclay Aclay + ksand (Abase − Asand )

kcrit = (4.3)
Abase
For simplified calculation purposes it is recommended to assume a single permeability (kcrit ). For
more advanced calculations it is recommended to use appropriate finite element software-tools (see
Chapter 6). Based on back-calculations the fracture-ratio (Aclay /Abase ) can be determined which is
required to reduce the tip resistance. The fracture-ratio is defined by Aclay /Abase , if this ratio equals
1, there is no crack requirement.
4.1.2 Suction below the plug

As stated previously, the suction and thus the amount of seepage flow just below the clay plug
determines the amount of the reduction of tip resistance and inner friction in the underlying sand.
This reduction is of interest to estimate the installation feasibility. It is expected that the suction just
below the clay plug will be a fraction of the applied suction within the caisson. Considering volume
continuity of the seepage through the plug, while the plug is not allowed to move, the following
equation is valid;
Sred S
qsand = qplug thus : ksand = kcrit (4.4)
γw s γw zplug
The boundary conditions of the 1D seepage problem is now defined, thus for each depth the parameter
kcrit can be solved according to equation 4.5.
! !
Scrit kcrit s
= (4.5)
Scrit + Splug ksand zplug
The subsequent subsections will be based on equation 4.5, whereas the critical permeability (kcrit )
can be back-calculated to the fracture-ratio (Aclay /Abase ).
4.1.3 Introducing the fracture ratio

Based on the calculations as presented in Chapter 2 the required suction is calculated. This suction
is required to install the suction caisson by inducing seepage flow in the underlying sand. The suction
should be applied to the underlying sand just below the clay plug and thus equals Sred . The true
suction in the caisson (S ) is limited by the uplift criteria for clay plug (equation 2.13) and thus the
critical permeability can be back-calculated according to equations 4.4 and 4.5. The calculated critical
permeability is then rewritten into a fracture-ratio, which is presented in Figure 4.3. A step-wise
procedure to determine the fracture ratio can be found in Appendix D.3.

Ac/Ab [-] Ac/Ab [-]

0,6 0,7 0,8 0,9 1,0 0,6 0,7 0,8 0,9 1,0
0,00 0,00
Houlsby Houlsby
1,00 1,00
NGI (HE) NGI (HE)
2,00 2,00
Senders (HE) Senders (HE)
3,00 3,00
4,00 4,00
5,00 5,00
6,00 6,00
7,00 7,00
8,00 8,00
9,00 9,00
10,00 10,00
(a) High permeable sand ksand = 1e−4 [m/s] (b) Lower permeable sand: ksand = 1e−5 [m/s]
Solid line: kclay = 1e−6 [m/s] Solid: kclay = 1e−6 [m/s]
Dashed line: kclay = 1e−10 [m/s] Dashed: kclay = 1e−10 [m/s]
Figure 4.3 Fracture ratios for high and lower permeable sands
From Figure 4.3 it can be noted that with increasing depth the ’required’ amount of fractures in
the clay is constant. When Ac /Ab equals 1, this implies that the clay plug is intact, which is the case
during the initial 3 meters of penetration. The permeability of the sand was taken as k = 1e−4 m/s.
The Houlsby-method shows some deviation compared with the CPT-based methods, which is likely
caused by the iterative calculation method of the soil resistances (see Chapter 2). For increasing
depth, all methods show approximately the same fracture requirements.
As can be found in Figure 4.3a, there is a limited change of fracture ratio for permeabilities in
the range of 1e−10 m/s up to 1e−6 m/s. These values represent a range of clayey soils and therefore
it can be concluded that the permeability of the clay is not governing for plug stability. This can be
validated by changing the sand permeability, which is presented in Figure 4.3b. It shows that for less
permeable sand the fracture requirement is lower, which indicates that the sand permeability governs
the fracture requirement.
4.1.4 Effect of layer thickness
As can be found in Figure 4.3, the Houlsby-method deviates for shallow penetration in the sand.
Therefore the Houlsby-method is not considered for investigation of the effect of layer thickness. In
order to evaluate the effect of layer thickness on the fracture ratio, the thickness of the overlying
clay layer has been varied for four cases (Figure 4.4). It can be noted that for increasing clay layer
thicknesses, the fracture requirement increases (the ratio Ac /Ab decreases). For the condition where
7 m clay overlies the sand, at least 15 - 20 % of cracks are required.
These values for fracture-ratio are not realistic in practice. In accordance to the previous experi-
mental results (Chapter 3) it was found that for thin clay layers the fracture ratio was approximate
1%. In order to indicate a fracture ratio of 15 %, an example of 10 m diameter can be taken. For
a fracture ratio of 15 % it holds that a circular strip-shaped crack of 0.78 m is required. From a
practical point of view this is not realistic for the installation condition of required seepage flow. It
can therefore be concluded that due to cracking only, it is highly unlikely that plug stability can be
maintained during installation.

Plug stability 41
Ac/Ab [-]
0,75 0,85 0,95 1,05 1,15 1,25
0
2
4
Penetration depth [m]

6
8
10 NGI (HE) - 1 m clay
NGI (HE) - 3 m clay
NGI (HE) - 7 m clay
14 Senders (HE) - 1 m clay
Senders (HE) - 5 m clay
20
Figure 4.4 Effect of layer thickness on the fracture-ratio requirement
4.1.5 Conclusions
Assessment of the plug stability can be done by introducing a case with increased permeability of
the clay plug and hence preventing uplift. The requirement of additional cracks can be expressed by
a fracture ratio, which equals unity for an intact clay plug. From volume continuity it is found that
the fracture ratio should be 5 - 20 % for respectively a clay layer thickness of 3 m to 7 m. For the
condition where seepage flows are required to install the caisson it can be concluded that these values
are considered to be unrealistic in practice to maintain a stable plug.
4.2 Installation with a moving plug

From the theory within the framework of this thesis it was found in Chapter 2 that the residual plug
heave will be lower for higher installation-rates. This conclusion was supported by basic experiments
on the plug response of layered soils (Chapter 3), which indicated a rate-dependency on the plug
response. For installation conditions when uplift can be accommodated (e.g. by a permanent stick-
up height) the rate-effects of the clay plug should be assessed. This section discusses the condition
of a moving plug, based on volume continuity.
If the resultant suction over the clay plug becomes higher than the uplift criteria for the clay plug,
the plug tends to move upwards. This upward movement will be dependent on the inflow of water
through the underlying sand to fill the water gap. With volume continuity the uplift-rate of the plug
can be quantified.
4.2.1 Principle of volume continuity

Based on volume continuity, the equations 4.6 - 4.7 are valid for the suction assisted penetration phase
according to Senders [2008] (see Figure 4.5). The author also mentions a system-volume, which is
related to the bulk-modulus of water and takes the compressibility of the de-pressured water into
account. However, this contribution is negligible and is therefore not considered here.
∆Vpump = ∆Vdisplaced + ∆Vseep,clay + ∆Vseep,sand (4.6)
∆Vdisplaced = Ai v∆t (4.7)

ΔVpump
Suction caisson
ΔVdisplaced
hc ΔVseep,clay Clay
hs ΔVseep,sand
Sand
Figure 4.5 Volumes during suction-assisted penetration
The displaced volume (∆Vdisplaced ) is related to the installation speed of the suction caisson (v ). The
seepage volume consists of two components;
Seepage through the clay plug (1D);

S − Sred
∆Vseep,clay = kcrit Ai ∆t (4.8)
γw hc
Seepage through the underlying sand (3D);

Sred
∆Vseep,sand = ksand Ai ∆t (4.9)
γw s
The seepage length for the clay plug equals the height of the plug (1D). For the underlying sand,
the seepage length is determined by equation 4.13 (Senders [2008]) is taken into account, since it is
assumed that penetration into the underlying sand has taken place. If the clay plug is intact (i.e.
not fractured), the permeability of the clay is low compared to the sand and can be neglected in
the volume continuity equations. For simplicity it is therefore assumed that there will be no seepage
through the (lower permeable) clay plug during uplift.
If the plug is assumed to be a solid seal, a relation with the installation-rate can be made.
Equations 4.10 - 4.13 are valid for Figure 4.6, where an upward movement is considered to be
positive, i.e. the velocity of the caisson (vcan ) is negative and the plug (vplug ) is taken positive.
It can be seen that the uplift-rate of the plug (vplug ) depends on the permeability of the underlying
sand, installation-rate (vcan ) and the level of suction within the caisson.
∆Vpump = ∆Vdispl,can + ∆Vdispl,plug + ∆Vseep,sand (4.10)
∆Vdispl,can = Aplug (−vcan + vplug )∆t (4.11)
∆Vdispl,plug = Aplug (qsand + vplug )∆t (4.12)
∆S2
∆Vseep,sand = ksand Abase ∆t (4.13)
γw s

Plug stability 43
ΔVpump
Suction caisson
ΔVdisp,can
Plug
hc ΔVdisp,plug Clay
hs ΔVseep,sand
Sand
Figure 4.6 Volume continuity during uplift
4.2.2 Suction below the plug
It is assumed that when the plug starts to move upwards, the pressure drop over the plug stays
constant (no seepage through clay plug). Since the pressure stays constant over the clay plug, the
applied pressure above the plug will be transferred just below the plug as the suction further increases.
As a result the plug moves upwards and suction is applied on the sand below the clay plug.
The plug can be schematized as a solid seal with downward pressure equal to Scrit,clay . If the
suction pressure ∆S1 is applied, the clay plug will move when ∆S1 > Scrit,clay . The pressure
below the clay plug is equal to the difference between the applied suction and critical uplift suction;
∆S2 = ∆S1 − Scrit,clay (Figure 4.7a). Figure 4.7b presents the development of suction within the
caisson, starting from initial phase with hydrostatic pressures (black line). If the water is extracted
within the caisson, the pressure drops with magnitude S, but with a reduced magnitude Sred just
below the plug.
P (kPa)
Suction caisson Suction

S caisson
ΔS1
Plug
Plug
ΔS2 Clay Clay
Sred
Sand αSred
d(m)
Sand
(a) Pressure drop over the clay layer (b) Pressure drop through clay and sand
Figure 4.7 Suction inside caisson with idealized pressure drop
The clay plug will move once the suction pressure exceeds the submerged weight and frictional
resistances of the clay plug (critical suction). From Figure 4.8 it can be noted that the components
of plug resistance consist of a self-weight and frictional-term. Once ∆S1 becomes lower than Scrit ,
a suction pressure below the clay plug develops (∆S2 ) which will induce a seepage flow. In the
subsequent description no contribution of reverse end bearing is incorporated. The reverse end
bearing is further discussed in Chapter 5.

200
Suction pressure [kPa]

S1
150
S2
100 S_uplift
50
Friction
0 Self-weight
0 200 400 600 800 1000
Time [s]
Figure 4.8 Contributions of plug uplift resistances
4.2.3 Uplift-rate of the plug

The uplift rate can be determined based on the volume continuity. In order to investigate the effects
of uplift-rate, a variation of penetration-rate is considered. From equations 4.10 - 4.13 it can be seen
that the uplift-rate of the plug (vplug ) depends on the permeability of the underlying sand and the
installation-rate of the caisson. This indicates the bilateral dependency of plug uplift, which involves
a coupled differential equation to be solved. However, by analytical elaboration it can be seen that
for higher installation-rates, the increment of displaced volume of the can (∆Vdisp,can ) reduces faster
if time-step ∆t is kept constant (equation 4.11). As a result the total pumped volume in time will
increase faster and thus installation is achieved quicker. Considering the components that contribute
to plug uplift, it can be found that a smaller time period corresponds to lower plug velocity. This is
in accordance to the theory found in Chapter 2 and experimental results in Chapter 3.
The total uplift can be determined from the total volumes as presented in the previous sections.
The difference in total volume of Vdisp,plug and Vseep is equal to the volume which represents the
uplift of the plug. The plug uplift is then calculated by dividing this volume by the base area. As
well there will be a contribution of elastic plug heave, due to displaced soil of the installed caisson.
Further details about elastic plug heave are presented in Appendix D.2.
4.2.4 Conclusions
Based on analytical analysis it is concluded that for slow installation (i.e. relatively low pump
flow rates) the total uplift is higher compared with fast installation. This is in accordance to the
theory found in Chapter 2 and experimental results in Chapter 3. The governing component which
determines the speed of uplift of the plug is the seepage flow of the underlying sand. It can therefore
be concluded that the permeability and applied suction are key parameters which influence the rate
of plug uplift. From a practical point of view, it can be remarked that when the caisson is halted
during installation, the under pressure below the clay plug is still applied and thus seepage flow is
induced, i.e. the plug will lift upwards. Once the pressures equate, the plug will finally move back
downwards due to self-weight.
4.3 Conclusions
If plug uplift has to be minimized, the plug stability can be assessed by introducing two extreme
cases; a stable plug and a moving plug during installation. In the first case, the permeability (and
thus seepage) is increased to some extent by cracking of the clay plug. Stability of a stable plug can

Plug stability 45
be assessed by introducing a fracture-ratio, which determines the crack-requirement for installation

without uplift. From Figure 4.9a it can be concluded that for clayey soils (k ≈ 1e−7 m/s), at least
10% of cracking is required to install the caisson (if tip reduction is required) in the underlying sand
with a stable plug. From a practical point of view this not realistic for the installation condition of
required seepage flow. It can therefore be concluded that due to cracking only it is highly unlikely
that plug stability can be maintained during installation.
Ac/Ab [-]
Ac/Ab [-] 0,75 0,85 0,95 1,05 1,15 1,25
0,6 0,7 0,8 0,9 1,0
0,00
0
Houlsby 2
1,00
NGI (HE)
2,00
4
Senders (HE)

3,00 6
4,00 8
5,00
NGI (HE) - 3 m clay
6,00 12 NGI (HE) - 5 m clay
NGI (HE) - 7 m clay
7,00 14 Senders (HE) - 1 m clay
Senders (HE) - 5 m clay
10,00 20
(a) Fracture-ratio for 3 m clay (b) Effect of layer thickness on fracture ratio
Figure 4.9 Fracture ratios
The effect of layer thickness of the overlying clay layer shows that for increasing thicknesses of
the clay layer the required amount of cracks also increases (Figure 4.9b). A maximum of 20 % was
calculated for a 7 m thick clay layer, which approximate remains constant for increasing depths.
Again a remark about the practical applicability can be made here, since tip reduction is required
and plug uplift cannot be accomodated.
When uplift can be accommodated by designing a permanent stick-up height of the caisson, the
relation between the plug velocity and the installation-rate should be assessed properly to prevent
excessive plug heave. Based on analytical analysis it is concluded that for slow installation the total
uplift is higher compared with fast installation. This is in accordance to the theory found in Chapter
2 and experimental results in Chapter 3. The governing component which determines the amount of
uplift of the plug is the permeability of the underlying sand and applied suction. From a practical
point of view, it can be remarked that when the caisson is halted during installation, the under pres-
sure below the clay plug is still applied and thus seepage flow is induced, i.e. the plug will lift upwards.
Furthermore it can be remarked that possible contributions of reverse end bearing are not imple-
mented in the analysis. This implies more conservative results, which should be recognized. The
effect of capacity due to reverse end bearing is discussed in Chapter 5.


Chapter 5
Modelling components of plug

uplift
This Chapter presents a method to model the plug uplift mechanism for the condition where sand is
overlaid by clay. In section 5.1 the mechanism for uplift of the clay plug is discussed. The next section
(5.2) focuses on the possibility of passive suction contributions (reverse end bearing) in layered soil
profiles. A generic model to describe plug uplift is presented in section 5.3. Finally the conclusions
are presented in section 5.4.
5.1 Mechanisms at the interface

A brief description of the breakout time related to the bottom in-situ time of an object (pipeline) is
presented by Roderick and Lubbad [1975]. This paper describes the bottom breakout mechanism in
detail, which can also be considered for the plug uplift mechanism. The mobilization of reverse end
bearing during ’breakout’ of the clay plug is shown in Figure 5.1b.
Clay Clay
Sand Sand
(a) Self-weight and shaft resistance (b) Contribution of reverse end bearing
Figure 5.1 Components of plug uplift
As soon as suction in the caisson is mobilized, the resultant force on the plug causes a motion of
the plug. This leads to a reduction in pore water pressures in the seabed beneath the plug relative to
the pressures in the surrounding sand. This gives rise to the suction force mobilized by the plug, which
could possibly be considered temporarily as extra capacity to withstand uplift (suction mobilization
in Figure 5.2b).
47
Suction mobilization
If the uplift force is less than required to cause immediate uplift (Qs,max ), the time required for uplift
depends on the force magnitude and seabed properties. The reduction of pore pressures in the seabed
under the object causes a flow of water from the surrounding material towards the low pressure area.
The rate of flow depends on the coefficient of permeability of the soil and the magnitude of pressure
difference. The suction force increases from zero to a maximum value.
Suction plateau
6628 STEEL CATENARY RISER TOUCHDOWN POINT INTERACTION MODELS 7
The suction plateau in Figure 5.2b is presented as plateau distance, where the displacement of the pipe
is shown on the horizontal axis. For breakout of the pipe (or plug), this distance represents a force
Initial Condition Soil Resistance Force per
equilibrium 1phase where the downward unit suction
Length,force
QU is maintained for a certain upward movement.
For constant installation-rates, the horizontal axis can be replaced by a time axis while maintaining
Backbone Curve
the suction curve. The plateau distance will then become Pipe/SoilaInteraction
time period where the suction remains
constant. This is interesting for installation practice,Curve because this time period can be regarded as
Previous Pipe/Soil
extra safety to ’delay’ uplift of the plug. Interaction Curve
Suction release Depth, z
As water moves Penetrate

to the low
Pipe pressure Soil Resistance
using Self zone, Force per
the pressure difference decreases and there is an accom-
2 Weight unit Length, QU
panying upward movement of the object. As the object movement becomes greater, the dissipation
of pressure differences and seabed deformations cause a loss of strength and uplift occurs (Suction
release in Figure 5.2b). Under further upward movement of the plug the suction force reduces from
its maximum to zero. If the uplift force is large or isInitial increased
Pipe due to high suction in the caisson,
plastic deformations might occur. This condition is Penetration
not expected for suction caisson installation,
OTC 16628 9
since the mobilized suction is gradually decreased during the installation phase.
Depth, z
Upliftper
Soil Resistance Force
Pipe Moves Upwards Resistance
3 unit Length, QU Force (kN) Suction Plateau
Maximum Analytical Model

Uplift Observed Soil Curve
Resistance
Force, Q S,MAX
Suction Release
Suction
Mobilisation
Depth, z
Break-out
Pipe Moves into Contact Soil Resistance Force per Displacement
(a) Upward
4 movement of pipe
with Soil unit Length, QU (b) Schematization of suction capacity
Figure 5.2 Pipe/soil interaction after Bridge et al. [2004]
Considering a layered soil profile, where sand is overlaid by clay, the theory for object breakout is
applicable to the underlying sand if the clay plug is considered to be impermeable. From literature it is
found that the assessment of suction capacity is based on dissipation of negative excess pore pressures
Depth, z
and thus is related to the permeability of the seabed (Gourvenec et al. [2009]). An assessment based
Pipe isofPushed
on the permeability Soil is
Further into sand
the underlying Resistance
thereforeForce per
proposed to investigate a possible contribution
5 Soil unit Length, QU
of passive suction capacity in sand. This is presented in the next section (section 5.2).
Pipe
Reload
Depth, z
Figure 2 – Illustration of Pipe/Soil Interaction

Modelling components of plug uplift 49
5.2 Reverse end bearing in sand

Within the framework of this thesis it was found that reverse end bearing is not considered as potential
contribution for installation layered soils (Chapter 2). Additionally, the experimental test results
(Chapter 3) and analytical analysis (Chapter 4) excluded possible contribution of passive suction
capacity. This section discusses the components of reverse end bearing in layered soils, especially for
conditions where sand is overlaid by clay. For homogeneous clays the benefit of caisson installation
with regards to plug stability, lies in their ability to resist uplift by generating significant resistance
due to passive suction capacity (in literature this phenomenon is defined by reverse end bearing).
This capacity is governed by the development of negative excess pore pressures (suction) within the
confined soil plug. If these passive suctions can be maintained, a significant uplift resistance is given
by reverse end bearing. Especially for large diameter caissons, the capacity due to reverse end bearing
could be governing, since the passive suction capacity is mobilized over the entire cross sectional base.
Water out Water out
Clay
Sand
(a) Penetration in top of clay layer (b) Penetration in centre of clay layer
Water out Water out
(c) Penetration nearly through clay layer (d) Penetration in underlying sand layer
Figure 5.3 Contribution of reverse end bearing for different penetration depths
Figure 5.3 presents a simplified overview of the mobilization of reverse end bearing in case of
penetration in layered soil profiles. The schematization indicates that for penetration through the
clay (Figure 5.3d), the contribution of reverse end bearing relies on the properties of the underlying
sand (e.g. permeability). For layered soils, when for example sand is overlaid by clay, it is uncertain
what the (remaining) capacity of reverse end bearing is and how long this can be considered. In order
to assess this uncertainty, first the reverse end bearing is evaluated for homogenous clay conditions.

Reverse end bearing capacity in clay
The generated capacity due to reverse end bearing in clay is calculated in a similar way as for ’regular’
end bearing capacity (Verruijt and van Baars [2007]), i.e.;
PREB = Nc su (5.1)
Where Nc is estimated by many authors and guidelines (DnV [1992]) and is assumed to be 9
for greater depths (see Appendix B). This value is based on Prandtl’s-theory for a strip loading
(Nc = 5.14) and extrapolated for circular footings and depth dependency. The factor is based on a
critical stress state and therefore neglects any consolidation effects. For the condition of layered soils
this is a critical aspect (for the underlying sand) and thus the consolidation time should be assessed.
Therefore the next subsection discusses the time dependency of reverse end bearing in the underlying
sand.
5.2.1 Time dependency of reverse end bearing

The reverse end bearing capacity in time relies on the dissipation of the negative excess pore pressures
in the seabed. For homogeneous clays the dissipation of pore pressures develops slowly and is similar
to a consolidation problem, as indicated in previous section. The time-dependency of the dissipation
during consolidation (or reverse end bearing) depends on the permeability of the clay and drainage
path, i.e. suction will dissipate faster if the flow through the seabed can be higher or the distance
is shorter. This is also stated by Huang et al. [2003], were the magnitude of suction to be sustained
greatly depends on the permeability of the subsurface soil and drainage paths. Higher permeable
soils will tend to decrease the suction faster under sustained loads.
Di
Suction caisson
pplug

Sand
REB = f(t,ksand)
Figure 5.4 Schematization of reverse end bearing during installation
From this it can be concluded that the reverse end bearing capacity is clearly related to the
permeability of the soil and drainage path. However, for clays, a dimensionless Nc -factor is taken to
estimate the capacity. The factor is based on total stress failure and should be treated with great
care, since it does not incorporate soil properties (e.g. permeability). The properties which determine
the capacity of reverse end bearing are the undrained shear strength su (magnitude of capacity) and
permeability of the clay (amount of time to maintain capacity). A schematization can be found in
Figure 5.4. Since the permeability of clays is very low, the total stress failure will intervene first
according to Prandtl’s-theory.
For layered soils (as presented in Figure 5.3d), the generation of negative excess pore pressures
will develop in the underlying sand. Therefore the permeability and seepage length of the underlying
sand should be considered in order to assess the time-effects.

Introducing the time lag
For clays, with permeabilities of at least 103 times lower than sand, the dissipation of the suction takes
substantially more time than for sand. The effects of permeability can be explained by introducing
the time lag 1 . The elaboration of the differential equation is presented in Appendix E.1 and a
definition is given by;
s seepage length
T = = [s] (5.2)
k permeability
From this equation it can be concluded that for lower permeabilities the time lag increases, thus
the dissipation of the suction takes longer. In order to put this in perspective, a simple example is
evaluated with regards to the typical sand and clay permeabilities. The time lag for this example is
shown in Figure 5.5, the range of normalized suction is typical for suction caisson installation.
7,00
6,00
Permeabilities
5,00 [m/s]
4,00
D/z [-]
1,00E-03
1,00E-04
3,00
1,00E-05
2,00 1,00E-06
1,00E-07
1,00 1,00E-08
0,00
3 months
1 month
1 day
1 week
1 hr
Time lag (T) [-]
Figure 5.5 Time lag for range of permeabilities
The horizontal time - axis in Figure 5.5 is an extended range compared to practical conditions
as the installation of suction caissons roughly is done in 1 - 6 hours, depending on geometry and
soil conditions. Each line represents a range of time - lag T and is depending on caisson dimensions
and penetration depth. It can be noted that for k = 1e−3 [m/s] the time-lag is in the range of
suction caisson installation, thus the pressures will equalize relatively quick. Reverse end bearing is
mobilized, but is maintained very short and capacity due to passive suction should not be considered.
However, for a permeability of k = 1e−5 [m/s], it can be concluded that the time required for
equalization is more than one day for the chosen range of dimensions. For a typical installation period
of 6 hours, the passive suction can be considered, since the time required for water inflow is lower
than a typical installation period.
For clayey soils (3 right curves) it can be seen that the range of time lag exceeds one week. Con-
sidering typically installation time periods, this implies that the time for dissipation of the negative
excess pore pressures (reverse end bearing) in these soils is more than the time period of installation.
1 The time required for water to flow to or from a point in a subsurface, until a desired degree of pressure equalization
is attained (Hvorslev [1951]. For this approach, the definition of flow lag would be more appropriate, since the delayed
inflow of water is considered.

Practical perspective
It should be remarked that this theoretical approach is based on purely homogeneous soils, which are
rarely found in practice. However, a distinction in the permeabilities of sand which might be favorable
to passive suction generation can be recognized in Figure 5.5. These values can be implemented for
the design phase in engineering practice by carefully determination of the permeability of the sandy
soil. This can be achieved by CPT-tests which include a dissipation test on the pore pressures. In
a dissipation test the porewater pressure change is obtained by recording the values of the pressure
against time during a pause in pushing and while the cone penetrometer is held stationary. As a
result the dissipation (contribution of reverse end bearing) can be quantified, which is of interest to
assess the plug stability. Additionally, oedometer tests or correlations with particle size distributions
can be used to determine (a range of) the in-situ hydraulic conductivity properties.
5.2.2 Conclusions
The plug uplift mechanism is schematized by using the theory for object breakout from the seafloor.
This theory is based on recommendations (DnV [1992]) for clay properties only and hence not applic-
able for sandy soils or layered soil profiles. However, from literature it was found that the assessment
of suction capacity is based on dissipation of negative excess pore pressures and thus is related to
the permeability of the seabed. An assessment based on seabed permeability is therefore proposed
to investigate a possible contribution of passive suction capacity in sand.
Reverse end bearing capacity in time relies on the dissipations of negative excess pore pressures
in time and is related to the soil permeability and drainage length. For clays, the permeability is low
and therefore total stress failure will intervene first during installations.
For sands, the possible capacity in time of pore pressure dissipation depends on the time required to
equalize the pressure to its initial value. This time requirement is defined by the ’time lag’ and can
be calculated for a range of caisson dimensions, pressures and permeabilities. This shows that for a
typical installation period of 1 - 6 hours, for permeabilities lower than 1e−5 m/s some reverse and
bearing capacity due to delayed inflow of water can be build up, which is interesting for installation
practice.
5.3 Modelling the plug uplift mechanism

Previous section decribed the time-effects related to contribution of reverse end bearing capacity in
layered soils. In order to further assess the contribution of reverse end bearing is these conditions, a
simplified model based on individual components is adopted to describe the uplift behaviour. This
section describes the plug uplift aspects by dividing the contributions two components; uplift by
suction acting on top of the clay plug and uplift due to seepage inflow below the clay plug. These
components are described in the first two paragraphs, followed by a combination of the components
in the third paragraph. This analysis presents a first approximation of the governing components
which determine plug uplift. It should be noted that this model is strongly simplified and is a set-up
to indicate the effects of individual components during uplift of the plug.

5.3.1 Uplift by suction only
In theory the plug uplift is linearly proportional to the applied suction, according to a spring model
(Figure 5.6). Similar to Hooke’s Law, elastic potential energy is stored as a result of deformation of
an elastic object, in this case plug uplift. The model assumes no boundary effects of underlying sand
and thus the plug movement depends only on the applied resultant force and spring stiffness.
Δs
PLUG
Figure 5.6 Spring-model for plug uplift
Once the plug is moving upwards due to the resultant upward force, the volume above the plug
will decrease. From Figure 5.7 it can be noted that V3a is lower than V2 due to the uplift of the
plug. Since a negative pressure is applied (suction) above the plug, a volume decrease of V2 to V3a
will increase the pressure above the plug. An increased pressure lowers the suction on the plug, thus
the underpressure is brought back to its initial value. This can be observed by a drop in suction
pressure during sudden uplift of the plug. As a result the resultant force (or suction) lowers and the
upward movement will reduce. In the next time-step the suction can be built up again, according to
the spring-model. A detailed flow-chart is presented in Appendix E.2.
SAbase
∆splug = (5.3)
k
k = f (su , D, γ 0 , z) (5.4)
Suction caisson
V1 V2 V3a
Plug
Plug V3b Clay
Sand
Figure 5.7 Volume above plug during installation
In order to account for this iterative process, the movement of the clay plug is defined by the
mechanical ’work’ definition; a force is said to do work when it acts on a body so that there is a
displacement of the point of application, in the direction of the force. For the application of a clay

plug, the force is obtained by the applied suction multiplied by cross area;
∆W (t) = S(t)Abase ∆splug (5.5)
The iterative process is then implemented by taking the increments in suction and volume to
calculate the generated work (W ). A small time-step should be taken to avoid large deviations of the
calculations as indicated in Figure 5.8;
∆W (t) = ∆P (t−1 )∆V (t−1 ) (5.6)
-0,09
-0,08
-0,07
-0,06
Uplift [m]
-0,05
-0,04 Δt = 9 s
-0,03 Δt = 15 s
-0,02 Δt = 11 s
-0,01
0
0 5000 10000 15000
Time [s]
Figure 5.8 Effect of large time-step for calculations
It can be noted that the general trend for uplift due to suction only, consists of an initial increase
in plug uplift. During installation (increasing time and depth) the increments for uplift deceases
because the incremental work applied to the plug decreases. Combining both equations give the
equation to describe the uplift of the plug by;
∆P (t−1 )∆V (t−1 )

∆splug = (5.7)
S(t)Abase
5.3.2 Uplift by seepage inflow

During uplift of the clay plug, the gap below the plug is filled with extracted water from the underlying
sand. As stated in the previous section, the plug uplift due to seepage is time-dependent. If the plug
is assumed to be impermeable, the total volume of water inflow during installation equals the residual
plug heave.
∆splug (t) S(t)

= qsand = −ksand (5.8)
dt γw s
Figure 5.9 shows a schematization of the plug uplift due to seepage inflow by a dashpot. Dashpots
are elements to model time-dependent behavior and are therefore applicable for uplift due to seepage
inflow. In contrary to a spring model, energy will dissipate as a result of deformation of the object.
This implies that there will be residual heave depending on the installation time.

Δs
PLUG
Figure 5.9 Dashpot-model for plug uplift
As indicated in section 5.1, the amount of plug heave is highly dependent on the soil permeability
and thus a range of permeabilities is chosen to assess the plug heave in Figure 5.10. The installation
time for this example was set to 2.5 hr with a constant pumping rate of 300 m3 /hr. It can be
noted that the plug uplift accelerates for increasing penetration in sand (and time). This is due
to increasing suction and seepage length, which results in a wide range of plug uplift for different
hydraulic conductivities. This indicates that the permeability of the underlying sand has great impact
on the residual plug heave.
-1,2
-1
k1 = 5e-4 m/s
-0,8
k2 = 1e-4 m/s
Uplift [m]
-0,6 k3 = 1e-5 m/s
-0,4
-0,2
0
0 2 4 6 8 10
Penetration in sand [m]
Figure 5.10 Plug heave due to seepage flow for a range of permeabilities
5.3.3 Combination of suction and seepage inflow

The total uplift of the plug can be described by combining both components of the uplift mechanism;
uplift by suction and uplift by seepage inflow. The question arises whether the components should
be placed in series or in parallel. Regarding the orientation of the plug, (suction on top of plug and
seepage below) the composition of the components seems to be in series according to Figure 5.11.
If it is assumed that the plug stays intact during uplift (sealing effect and rigid plug), the dis-
placements of both components have to be equal according to Figure 5.11. This also implies that a
possible higher suction on top of the plug (which will cause a certain uplift of the plug) is restricted
by the seepage inflow below the plug. This principle relies on dissipation of pore pressures in time
and is considered to be similar to the reverse end bearing principle. However, as indicated in previous
sections, it should be remarked that this schematization is strongly idealized. The model considered
a residual plug heave due to the deformed dashpot, which is not realistic in practice. Once the
pressures equalize, the plug will settle due to self-weight of the plug.

Δs
PLUG
Figure 5.11 Combination in series to model plug uplift
Modelling reverse end bearing
The reverse end bearing capacity can be schematized by the approach in which the components of
suction only and seepage inflow are included. If is assumed that the clay plug will be an impermeable
sealed plug and not subjected to elastic deformations, a theoretical assessment of the reverse end
bearing can be made. As stated in the previous section, the displacement of the plug depends on
both suction (applied on top of the plug) and seepage flow (below the plug). This is presented
in Figure 5.12, where it can be noted that for suction only the uplift is more or less constant for
increasing depth, whereas the contribution of seepage inflow below the plug slowly increases for
increasing depth.
It should be remarked that the chosen value of permeability is highly dependent on the calculated
plug uplift (as indicated in Chapter 4) in Figure 5.12. The amount of uplift can be considered as
indication of the contributions. The contour-line shows an indication of the maximum expected plug
heave.
-0,40
-0,35 s,suction only

s,seepage only
-0,30 s,suc+seep (contour)
-0,25
Uplift [m]
-0,20
A B
-0,15
-0,10
-0,05
0,00
0 2 4 6 8 10
Pentration in sand [m]
Figure 5.12 Components of plug uplift
The starting point of both ’contributions’ is around 2 m penetration. The self-weight and shaft
resistance of the plug are incorporated for the calculations, by calculating the suction requirement
for plug uplift according to Senders [2008]. It can be remarked that the schematized contribution

of reverse end bearing is mobilized when plug uplift initiates. This is accordance to the break-out
resistance of Bridge et al. [2004] and the uplift-resistance of Gourvenec et al. [2009], where a small
displacement is required to mobilize resistance against uplift. Considering both contributions in
Figure 5.12 of plug uplift, a distinction between 2 phases can be made;
Phase A
The theoretical uplift due to suction only is higher than that for seepage inflow. Since both contri-
butions are coupled (impermeable and rigid plug), the actual uplift depends on the contribution of
the seepage inflow, which is lower.
Phase B
The theoretical uplift caused by the seepage inflow is larger than that for suction only. The former
will be the governing contribution to the plug heave. The plug is ’pushed’ upwards due to the inflow
of water just below the plug.
The most interesting condition is Phase A, where a downward resistance (due to resitrictions) is
generated as long as the uplift due to seepage is lower than that for suction only. Figure 5.13
presents a plot of the differential uplift of both components for Phase A. It can be noted that Figure
5.13 shows similarities with the passive suction model according to Bridge et al. [2004] in Figure 5.2b.
The suction plateau is negligible in this example, thus the capacity consists of mobilization followed
by release of the suction capacity. The total time required to mobilize the capacity in this example
is 40 minutes.
0,035
0,030
0,025
Δs [m]
0,020
0,015
0,010
0,005
0,000
0,0 2,0 4,0 6,0 8,0 10,0
Penetration in sand [m]
Figure 5.13 Difference in uplift with penetration depth
It should be remarked that Figure 5.13 only indicates duration of negative excess pore pressure
generation (reverse end bearing), based on the difference in uplift by two components acting on
the plug. The magnitude of capacity of reverse end bearing is not indicated by this figure. The
model assumes constant penetration-rate, thus effect of halted penetration are not considered. If the
penetration is stopped, uplift will continue due to seepage inflow below the plug untill equilibrium of
pressures is reached. Hereafter the plug will settle due to self-weight of the plug.
5.3.4 Conclusions
With the implementation of a spring-dashpot model the components which determine uplift of the
plug can be described. The applied suction above the plug will initiate movement of the plug, but
this will be restricted by static components (e.g. skin friction and self-weight) and inflow of water
below the plug. For the condition when the applied suction on top of the clay plug is higher than
seepage inflow below the plug, an indication of negative excess pore pressures is presented.

5.4 Conclusions
Considering the passive suction capacity successively, the phases of Suction mobilization, - plateau
and - release are distinguished. Disregarding the soil type, it was found that the magnitude of
resistance (Suction mobilization) and duration (Suction plateau) depends on the dissipation of neg-
ative excess pore pressures in time. This can be related to the permeability of the soil and shows
that for clayey/silty soils, the time for dissipation exceeds a typical installation time period by a
large extent. The magnitude of the maximum uplift resistance and the duration of the resistance
(time period of suction plateau) are believed to be most important for assessment of plug uplift fail-
ure. Significant contributions of plug uplift resistance can be considered for installations of caissons
in clayey soils. In case of installations in layered soils, where sand is overlaid by clay, the time to
maintain reverse end bearing capacity is expected to be lower since the permeability of sand is higher.
The possible capacity of excess pore pressure dissipation depends on the time required to equal-
ize the pressure to its initial value. This time requirement is defined by the time lag and can be
calculated for a range of caisson dimension, pressures, permeabilities and drainage length. This
shows that for a typical installation period of 1 - 6 hours and for permeabilities higher than 1e−5
m/s some capacity due to delayed inflow of water can be considered. It should be remarked that the
chosen range of permeability has a great impact on the time requirement for dissipation.
With the implementation of a spring-dashpot model the components which determine uplift of the
plug can be indicated. If the applied suction above the plug will initiate upward movement of the plug,
this will be restricted by the inflow of water below the plug. Besides the skin friction and self-weight
of the plug are static components, which will settle the plug after equalization of the pressures. Mo-
bilization of reverse end bearing can be schematized for cases when the uplift by suction only exceeds
that of seepage inflow (phase A in Figure 5.12). The time range of the generated capacity is based
on the condition where the theoretical uplift due to only suction is higher than that for seepage inflow.
The time requirement to maintain reverse end bearing capacity is greatly dependent on the chosen
permeability and drainage length. For more permeable soils the seepage flow increases rapidly, which
limits the time-period for reverse end bearing generation. As a result of this, it can be concluded
that plug heave will be higher for more permeable soils.

Chapter 6
Verification of the theory
This Chapter presents an analysis of two practical cases which are favourable to plug uplift, accom-
panied by a nummerical calculation on plug uplift. The first case, Case A, is published by Alhayari
et al. [1999] and discusses the anchors of the Curlew FPSO-project in section 6.1. The second case,
Case B, was published by Tjelta et al. [1986] and discusses the experimental program for installation
of the Gullfaks C-platform in section 6.2. Both cases include sufficient key-data to do back-analysis
for plug uplift assessment. Section 6.3 presents numerical verification of the theory on plug uplift
using finite element software. Finally the conclusions are presented in 6.4.
6.1 Case A: Curlew FPSO

For the Case A-site a group of three suction caissons were installed in the North Sea, where the soil
conditions are typically characterized by dense sand overlaid by less permeable soils. In this case the
dense sand is overlaid by clayey, silty sands, which is assumed to prevent seepage flow to develop
during the suction phase. The key-data of the three anchors are presented in Table 6.1 and Appendix
F.1.
Table 6.1 Key-data for Case A
Anchor Diameter Wall Penetration Total sub- Depth dense Depth very
[m] thickness depth [m] merged weight sand [m] dense sand
[m] [kN] [m]
1 7 0.03 9 600 4.0 6.8
2 7 0.03 9 600 4.0 6.8
3 7 0.03 9 600 4.7 8.1
In the design, no reduction in penetration resistance due to seepage is accounted due to high
concentrations of silt and clay. These layers may prevent the flow of water around the skirts during
the suction phase. In case of clean sands, this flow of water significantly reduces the skirt penetration
during suction. As can be found in Table 6.1, the soil consists of a loose sand up to 4 m (4.7 for
anchor 3), which overlies more dense sand up to 6.8 m (8.1 m for anchor 3). The very dense sand is
found below 6.8 m (8.1 m for anchor 3). An interpolated plot of the presented cone resistances can
be found in Figure 6.1.
Estimated resistances
In order to estimate the penetration resistances for the soil conditions of anchors 1 - 3, some reference-
calculations were done of historical data with similar soil conditions, e.g. dense sands in the North
Sea. The purpose of these reference calculations is to estimate the frictional terms (Kf ) for installation
59
qc [kPa]
0 5000 10000 15000 20000 25000 30000 35000
0
2
1&2
3

4
8
3
9
10
Figure 6.1 Interpolated cone resistance plot of soil profiles for anchors 1,2 & 3
of anchors 1 - 3. The total penetration resistance (Qtot ) consists of tip bearing and side friction and
is calculated according to;
Qtot = Qtip + Qskirt (6.1)
In the sand the tip resistance and side resistances are calculated as follows;
qtip = Kt qc and qskirt = Kf σv0 rtan(φ0 ) (6.2)
From the reference-data provided a good fit was found by using a Kt value of 0.2 and a lateral
earth pressure coefficient, Kf , varying from 0.5 to 2.0 dependent if the soil is layered with loose
silty sand (0.5) or dense sand (2.0). These coefficients are based on a best-fit approach of measured
suction and predicted suction of the reference-data. The skirt resistances (inside and outside) have
been adjusted to meet the best-fit solution for different layering of the soil. Based on these reference
calculations, a set of empirical coefficient, Kt and Kf , are used for calculation of the skirt penetration
resistance of anchors 1-3.
Table 6.2 Soil-data for calculations
Anchor 1 & 2 Anchor 3

Soil Profile qc [MPa] Depth [m] Kf [-] Depth [m] Kf [-]
Loose silty sand <5 0-4 0.5 0 - 4.7 0.5
Dense sand 5 - 20 4 - 6.8 1.0 4.7 - 8.1 1.0
Very dense sand > 20 6.8 - 9 2.0 8.1 - 9.5 2.0
Estimated suction
The estimated suction is based on the reference parameters and assumes no tip reduction due to
seepage flow. Thus the estimated suction is calculated in a similar way as calculations for installation
in homogeneous clay, where the suction causes a pressure differential across the top plate of the
caisson. This results effectively in an additional vertical load equal to the suction times the plan area
of the caisson.

Verification of the theory 61
P/ywD [-]
0 2 4 6 8 10
0,00
0,20
0,40
0,60
z/D [-]
0,80
1&2
1,00
1,20
3
1,40
1,60
Figure 6.2 Predicted suctions for installation of anchors 1 - 3
Figure 6.2 presents the estimated suction requirement for installation of anchors 1 - 3. It can be
noted that the suction requirement increases for the deeper dense sands (z/D > 0.95 for anchor 1 &
2 and z/D > 1.15 for anchor 3). The normalized suction of P/γw D = 8 corresponds to a suction of
515 kPa.
6.1.1 Measured suction

From the three anchors considered only a single suction measurement plot was available to compare
the predicted suction with the measured suction; i.e. only for anchor 3 the suction measurement was
available. From cone resistance data (Figure 6.3a) it can be noted that the input for the qc -values
are well determined. These input-values are required to estimate the resistances for penetration.
Cone Resistance [kPa]

0 10000 20000 30000 40000 0 10000 20000 30000 40000
0 0 0 0
1 CPT-data 1
1 1
qc-input Measured Suction
Kt=0.2 2 2
2 2 S,red
CPT-data Kf=0.5
3 3
3 3
4 4
4 4
5 5
5 5
6 6
6 Kt=0.2 6
Kf=1.0 7 7
7 7
8 8
8 8
Kt=0.2
9 9
9 Kf=2.0 9
10 10
10 10 0 100 200 300 400
Suction [kPa]
(a) Cone resistance (b) Measured suction
Figure 6.3 Measured versus estimated values for installation
Figure 6.3b presents the measured suction from the installation of anchor 3 by the black solid
line. The red line represents the estimated suction according to the qc -input in Figure 6.3a and Kt ,
Kf -values from Table 6.2. It can be noted that the measured suction is higher than the estimated
suction (red curve). This deviation is explained in the following paragraphs, where the water injection

devices and cross ties are discussed. First a best-fit back calculation of the suction is obtained for
further analysis.
Best-fit back calculated suction
Disregarding both effects stated above, a best-fit back-calculation can be done based on the Kt and
Kf -values (Figure 6.4). Figure 6.4a show that a reasonable approximation of the measured suction
can be obtained. The corresponding values for Kt and Kf are shown in Figure 6.4b. From the
latter figure it can be concluded that the input frictional term is underestimated for layers with high
qc -values.
Cone Resistance [kPa] Kp and Kf [-]

0 0,5 1 1,5 2 2,5
0 10000 20000 30000 40000
0
0 0
Input - Kf
1 CPT-data 1 1 Input - Kp
Measured Suction Back-fit - Kf
2 2 2
S,best fit Best-fit - Kp
3 S,red 3 3

4 4 4
5 5
5
6 6
6
7 7
7
8 8
8
9 9
9
10 10
0 100 200 300 400
Suction [kPa] 10
(a) Measured and back-calculated suction (b) Back-calculated shaft- and tip parameters
Figure 6.4 Back-calculations with best-fit approach
At depth 8 m the water injection devices might have reduced the tip resistance and thus Kp could
have equal zero. Even without tip resistance, still an overestimation of the suction is calculated at
a depth of 8 m, as indicated in Figure 6.5a. This could imply that plug uplift occurred, since 1 m
of penetration was achieved with constant 300 kPa suction. Verification of the possible plug uplift is
presented in subsection 6.1.2.
Cross ties penetration
The geometry of the anchors includes a 3 m high cross tie (lug stiffening), which is connected just
above halfway of the skirt. Once the anchor penetrated the soil at a depth of 4.5 m, the cross
tie contacts the soil and the penetration resistance will increase. Thus an increase of resistance is
expected at 4.5 m penetration, which can also be seen in Figure 6.4a and Figure 6.4b.
Water injection device
As the required and allowable suction were quite high and it was expected that the top silty sand
layer would act as a seal, it was decided to install water injection devices at the skirt tip. The
main advantage of these devices is that the tip resistance could be reduced by the outflow of water
at the skirt tip, which will liquefy the very dense sand and hence reduce the tip resistance during
installation. Since no data is available regarding the time or penetration depth of the injection, it is
difficult to determine the impact of the water injection.
However from Figure 6.4a it can be found that the measured suction slightly reduces, while
the estimated suction (and thus soil resistances) supposed to increase for penetration higher than

6.5 - 9 m. It is therefore assumed that the tip resistance reduces towards 0 from a penetration of
7 m (Figure 6.4b). This shows good agreement up to a penetration of 8 m, however beyond this
depth the prediction and back-calculation both overestimated the measurements. The reduction of
tip reduction could be caused by plug uplift or by the water injection devices at the skirt tip. These
effect are further discussed in subsection 6.1.2.
6.1.2 Plug uplift back-calculations

From the suction predictions it was assumed that no seepage flow will develop due to the less per-
meable top layers of silty, clayey sand. According to the plug uplift theory presented by Senders
[2008], a back-calculation can be done to check if this assumption is valid. For a clay plug the
equation is given by equation 6.3 and excludes contribution of reverse end bearing.

0 4
Pplug = γ + αsu zplug (6.3)
D
The shear term is valid for (remoulded) clay, but for this case replaced by shear resistance in
sand, i.e.;
fskirt = Kf σv0 rtan(φ0 ) (6.4)
In order to determine the uplift potential in a conservative way, the parameter Kf is taken as
2.0, which implies an overestimation of the shear resistance. Additionally, no contribution of reverse
end bearing is considered. The uplift potential is thus conservative since a higher uplift pressure is
required. A conservative back-calculation is chosen to ensure the most reliable back-calculation. The
exact plug thickness is arbitrary and therefore three interface possibilities for plug lift off are shown
in Figure 6.5. Only the top layer (0 - 4.7 m) consists of silty clayey sand and is therefore assumed to
be most favourable to plug uplift.
Cone resistance [kPa]

0 10000 20000 30000 40000
0
2 Kp=0.2
Kf=0.5
3
6 Kp=0.2
Kf=1.0
7
CPT-data
8 Plug1 Kp=0.2
9 Plug2 Kf=2.0
Plug3
10
Figure 6.5 Three potential plugs for uplift
Possible plugs 1 and 2 will generate resistance against plug uplift due to the underlying cohesive
soil. In general the total resistance against uplift consists of self-weight, shear resistance and passive
suction capacity (reverse end bearing). The latter contribution depends on the hydraulic conductivity
of the underlying soil and is higher for low-permeable soils. Considering the plug which is the most

favourable to uplift, the third plug (thickness 4.7 m) is taken for back-calculations.
Table 6.3 Potential plug dimensions for lift off
Plug No. Thickness [m] Pplug [kPa]

1 0-2 30.6
2 0 - 2.5 42.1
3 0 - 4.7 110.7
Figure 6.6 presents the criteria for plug uplift according to Senders [2008]. It can be seen that for
penetration more than 4.7 m the plug is favorable for uplift. The exceeding of the plug uplift criteria
is substantial, a further assessment of this possibility is presented in the subsequent paragraphs.

0 10000 20000 30000 40000
0 0
1 1
CPT-data
2 Measured Suction 2
S,best fit
3 3
P,plug
4 4
5 5
6 6
7 7
8 8
9 9
10 10
0 100 200 300 400
Suction [kPa]
Figure 6.6 Implementation of plug uplift for plug 3 (z = 4.7 m), without contribution of reverse end bearing
Analysing parameters
Soil parameters are determined prior to the design of the suction caisson. The most interesting
parameter for plug stability analyses is the permeability of the layered soil. With the use of this
parameter, the fracture requirement and possible plug heave (Chapter 4) and contribution of reverse
end bearing (Chapter 5) can be analysed. From the provided report it was found that an estimate of
the permeabilities was found by correlations to the grain size distributions. Based upon the report,
the maximum recommended soil permeabilites are;
Loose to medium dense silty sand: k = 1 x 10−5 [m/s]

Dense silty sand: k = 1 x 10−6 [m/s]
Fracture requirement
From the theory in the previous chapters, a theoretical fracture-ratio can be back-calculated for this
case. With rough estimations, it can be found that the fracture requirement varies between 25% for
shallow penetrations, which decreases towards 15% for target depth (Figure 6.7a). In order to put
this in perspective, a circular strip fracture of 0.9 m to 0.6 m is required to install without uplift.
This is supposed to be not realistic in practice, thus it can be concluded that cracking likely did not
occur.

Ac/Ab [-] Uplift [m]

0,6 0,7 0,7 0,8 0,8 0,9 0,9 1,0 1,0
0,00 0 0,005 0,01 0,015 0,02
1,00 NGI (HE) 0
2,00 Senders (HE) 2

3,00
4,00
4
5,00
6
6,00
8
7,00
8,00 10
9,00
12
10,00
(a) Fracture requirement for Case A (b) Plug heave estimate for Case A
Figure 6.7 Back-calculations for Case A
Plug heave
An estimate of the possible plug heave can be given by the total inflow of water below the sealing plug,
once the differential pressure over this plug exceeded the self-weight, shaft resistance and reverse end
bearing of the plug. For this example, no contribution of reverse end bearing is taken into account,
since this will give more conservative results for the heave estimate. With the given permeabilities,
suction requirement and installation time, an estimate of the plug heave can be made according to;
S
q = ksand (6.5)
γw s
∆splug = q∆t (6.6)
Where the time-step (∆t) depends on the installation time of the suction caisson. Since the
permeability of the soil is relatively low (for sandy soils) due to some silt fractions, it can be found
that the plug heave estimate is negligible (order of centimeters) in Figure 6.7b. From the published
data it was found that installation of the caisson was executed with a rate of 10 m/hr, thus the total
installation time is approximate 1 hour. Back-calculations of the estimated plug heave indicate that
the plug heave is very limited.
Reverse end bearing
The theory on reverse end bearing relies generally on the dissipation of negative excess pore pressures.
Since it is found that the permeability of the sand is relative low (for sandy soils), there might be
a contribution of reverse end bearing against plug uplift. This contribution is based on the delay of
water inflow, once the plug tends to lift off. The time requirement of pressure equalization depends
the time frame of the additional ’safety’.
With the estimated permeabilities of the soil strata, the time lag for this case can be determined.
Figure 6.8 presents the design permeabilities at a penetration depth which is favorable for plug uplift.
It can be seen that within the range of design permeabilities a time lag of 1 day is exceeded.
From the published data it was found that installation of the caisson was executed with a rate
of 10 m/hr, thus the total installation time is approximate 1 hour. Thus it can be concluded that
a contribution of reverse end bearing (delay of water backflow) theoretically could be relied upon.

7,00
Permeabilities
6,00
[m/s]
5,00 1,00E-03
4,00 1,00E-04
D/z [-]
1,00E-05
3,00
1,00E-06
2,00
1,00E-07
1,00 1,00E-08
0,00 CASE A
3 months
1 month
1 day
1 week
1 hr
Time lag (T) [-]
Figure 6.8 Time lage for Case A
Additionally, the water-injection-device was connected to the vessel’s fire hose system. Therefore
limited effects on pore pressure dissipation (reverse end bearing) are expected due to the relatively
low pressure. However it is unknown what the effects or influence of the water injection devices was
on the reverse end bearing development. Since the latter relies on dissipation of pore pressures, while
near the tip ’extra’ water is foreseen, the reverse end bearing might have been decreased relatively
quickly.
Explanation of pressure drop
From the case data provided, no uplift was monitored and/or mentioned, thus another phenomenon
might have occurred which clarifies the pressure drop at penetration of 8 m. Figure 6.9 presents the
contributions of plug uplift according to Senders [2008], which are self-weight (Pweight ) and shaft
friction (Pf riction ). In addition the contribution of reverse end bearing (PREB ) is presented, based
on an Nc -value of 7.5. This value is based on best-fit analysis of the total plug uplift criteria (Pplug ).
Since the installation time of the caisson was relative short (1 hour), a contribution of reverse end
bearing could be applicable. It can be noted that the total plug uplift resistance now increases to a
value where the pressure drop in the caisson was encountered. This would imply that at the point
of pressure drop, the seepage flow was induced by the uplift of the plug. Initially this causes seepage
flow and thus reduction of the soil resistance near the tip. Simultaniously it causes limited uplift,
but this will tend to increase in time. It should be remarked that this principle is only valid when
it is assumed that the water injection devices near the tip have limited impact on the dissipation of
negative excess pore pressures; i.e. suction within the centre of the clay plug can still be maintained.
Limitations in practice
From a practical point of view some limitations to contribution of reverse end bearing can be men-
tioned. In the first place it can be remarked that during installation the soil skeleton close to the
skirt slightly changes due to penetration of the skirt. It should be noted that this may change the
density of the sand and thus affect the flow characteristics. Secondly, it can be remarked that in
cases of local shallow gas reservoirs, the dissipation of negative pore pressures could develop quicker
and thus no contribution should be taken into account.
6.1.3 Conclusions
In order to install the anchors in challenging conditions with very dense sands overlaid by a less
permeable layer. Since limited seepage flow was expected, water injection devices were implemented
as mitigating measure for successful installation. The installation resistances are estimated according
to reference-data for similar soil conditions. From the measured data it can be concluded that the


0 10000 20000 30000 40000
0 0
CPT-data
1 Measured Suction 1
S,best fit
2 P,plug 2
P,REB
3 P,weight 3

P,friction
4 4
5 5
6 6
7 7
8 8
9 9
10 10
0 100 200 300 400
Suction [kPa]
Figure 6.9 Implementation of plug uplift, divided by components
measured suction is partly in accordance with the estimated suction. Because of the implementation
of water injection devices and cross tie penetration, the measured suction deviates somewhat from
the predictions.
Disregarding these effects, still the potential for plug uplift failure can be assessed by calculations
according to Senders [2008]. This showed that for conservative calculations, the estimated plug heave
was limited (negligible), which is in accordance to the observations. However, interpretation of the
measured suction indicated a pressure drop at 8 m penetration depth, which can be clarified by the
contribution of reverse end bearing. This shows a reasonable agreement based on an Nc -value of 7.5.
The contribution of reverse end bearing was justified by the high installation-rate of 10 m/hr and
assuming limited effect of the water injection devices.

6.2 Case B: Gullfaks C

This case is based on the historical large-scale penetration test at a deep water-site by Tjelta et al.
[1986]. The Gullfaks C platform installation was scheduled in 1989, being the largest and heaviest
offshore concrete structure ever installed. During the foundation design phase it became evident
that geotechnical challagnes were expected. Therefore it was decided to execute a large scale-test to
improve the confidence in the predicted soil response and to clarify uncertain aspects. Key input-data
to perform back-calculations are presented in Table 6.4 and Appendix F.2.
Table 6.4 Key-data from Case B anchor installation
Diameter [m] Wall thickness [m] Penetration depth [m] Total submerged weight [kN]
6.5 0.035 22 3532
Water injection devices were implemented to ensure tip reduction in the dense sands, which were
expected at 18 m depth (Figure 6.10a). Figure 6.10b presents the measured tip resistance, which
are compared with back-calculated resistances of the soil profile presented in Figure 6.10a. A similar
approach as for Case A is used for the calculations, which shows good agreement with the measured
resistance.
Penetration Force [kN]

0 5000 10000 15000
0
5
Penetration Depth [m]
10
15
20
BACK
CALCULATED
RESISTANCE
25
(a) Soil profile (b) Measured and back-calculations resistance
Figure 6.10 Back-calculations with best-fit approach
6.2.1 Measured suction

Based on the resistances for penetration, one can estimate the suction requirement for installation.
If no tip- and shaft reduction due to seepage can be considered, the suction requirement is calculated
according to;
Qtot − W 0
Sreq = (6.7)
Ain

However, due to the use of water injection devices the tip resistance should be reduced to a
certain extent because of the increased pore pressures (and hence reduces effective sresses). Figure
6.11 presents the measured suction during installation of the caisson up to 22 m penetration depth.
The red curve shows the required suction if there is no reduction due to seepage and/or injection flow.
The dashed curve shows the required suction for reduced tip- and inner resistance (best-fit curve).
This reduction of resistance can be caused by the water injection devices, seepage flow through or
along the plug or increased permeability due to deformed soil or uplift of the clay plug.
Suction [kPa]
0 100 200 300 400 500
0
S,measured
S,no red
S,red
5
Water injection
device turned on
10
15
20
25
Figure 6.11 Measured and back-calculated suction (with and without tip reduction)
At a penetration of 9.8 m the water injection devices were switched on and from this point the
back-calculated suction is based on reduced tip and inner resistances. A best-fit suction requirement
was obtained where the tip and inner resistances were assumed to reduce 50 %, which is in accordance
to the theory presented in Figure 2.10a. The reduction of the inner resistance might be justified by
penetration of the skirt, which might affect the flow characteristics. From Figure 6.11 it can be noted
that the best-fit curve this shows good agreement with the measured suction.
6.2.2 Plug uplift back-calculations

Uplift is expected when the applied suction exceeds the plug uplift resistance, which is the case at
penetration depth of 20 m (Figure 6.12). Plug uplift (or heave) was also mentioned in the corres-
ponding paper of Tjelta et al. [1986] and estimated to be 1 m; 0.6 m due to displaced soil and 0.4
m due to applied suction. In order to assess the plug uplift potential a back-calculation of the plug
heave can be made with the use of installation time which is presented in the paper.
The confined sand layer (9.8 - 17 m) is very silty, clayey with a clay content of about 10 %
(Appendix F.2) and is therefore assumed to be rather cohesive and less permeable. In fact the
dense sand at 18 m depth is overlaid by a cohesive plug of thickness 17 meter. This assumption is
incorporated for the plug contour line in Figure 6.12.
Considering the stratification of the soil conditions, there’s a potential for plug uplift. In order
to assess the plug uplift criteria, the plug uplift contour is plotted together with the measured and
estimated suctions in Figure 6.12. The contour line in Figure 6.12a includes the possible component

of reverse end bearing (Supl,REB ). From this it can be noted that the contour line does not cross
the measured suction and therefore might be neglected. This observation is in accordance with the
relative low installation-rate (1 m/hr), which is not benificial for passive suction capacity.
Suction [kPa] Suction [kPa]

0 100 200 300 400 500 0 100 200 300 400 500
0 0
S,measured
S,red
5 S,upl,total 5
S,upl,weight S,measured
S,upl,fric
S,upl,REB S,red
S,upl,total

10 10 S,upl,weight
S,upl,fric
15 15
Possible uplift
20 20
25 25
(a) With reverse end bearing (b) Without reverse end bearing
Figure 6.12 Contributions of uplift components
This is confirmed by Figure 6.12b, where the contour line according to Senders [2008] shows good
agreement with the encountered uplift. The contribution of reverse end bearing is excluded in Figure
6.12b, where only the contributions of self-weight (Supl,weight ) and shaft friction (Supl,f ric ) according
to the theory of Senders [2008] are presented.
Fracture requirement
From the subsequent theory in the previous chapters, a theoretical fracture-ratio can be back-
calculated for this case. With rough estimations, it can be found that the fracture requirement
is around 40%. In order to put this in perspective, a circular strip fracture of 1.5 m is required to
install without uplift. This is supposed to be not realistic in practice, thus it can be concluded that
cracking likely did not occur.
Plug heave
Back-calculations rely on the inflow of water in a defined time period and depend on permeability
of the underlying sand and the applied suction. The latter can be directly related to the estimated
suction and the suction to lift off the plug. However, the permeability of the sand is not defined
and therefore an accepted range of sand permeability has been chosen. Since the underlying sand
(20 - 24 m) is more silty, clayey, the permeability was chosen to be half of the permeability of the
overlying dense sand (18 - 20 m).
An estimate of possible plug heave can be made according to the uplift theory of Senders [2008].
The paper of Tjelta et al. [1986] presented a plug heave of 0.4 m at target depth of 22.1 m, which
is also in the range of solutions in Figure 6.13. The paper stated that plug heave was observed (by
bottom clearance devices) in the last penetration stage, which is in accordance to the assumption

that plug heave occurs when the caisson penetrates the sand at 18 m. It can be remarked that the
variation in sand permeability leads to substantial differences in estimated plug heave. A range of
permeabilities for sand (5e−4 - 1e−4 [m/s]) is chosen to determine possible plug heave and showed
that the range of heave is 0.15 - 0.75 m for this case.
Heave [m]
0 0,5 1
18
Observed plug heave
18,5
s,plug1 | k = 5e-4 m/s

19 s,plug2 | k = 2e-4 m/s
19,5 s,plug3 | k = 1e-4 m/s
20
20,5
21
21,5
22
Figure 6.13 Estimated heave for range of permeabilities
However, if the range of permeabilities is chosen to decrease by a factor of 10, the plug heave
estimate will also decrease by a factor of 10 and becomes 0.02 - 0.07 m. This clearly indicates the
dependency of permeability and drainage length for plug heave. The installation time (close to
penetration) also has influence on the plug heave estimate. If the time requirement for installation
is increased by a factor of 2, the estimated plug heave will also increase by a factor of 2 and thus
implies a linear dependency. Both examples demonstrate that plug heave is highly dependent on
time-effects, i.e. permeability of the underlying sand and installation time.
6.2.3 Conclusions
Case B was analyzed by back-calculations of the suction requirement for reduced inner and tip
resistances. From the measured suction it can be concluded that the tip- and inner resistance were
reduced during installation, however it is uncertain whether this is caused by the water injection
devices or seepage flow due to plug uplift.
Considering the presented data of the paper Tjelta et al. [1986] plug heave of 0.4 m was encountered
during installation. Regarding the soil profile, a potential plug for uplift failure was determined and
back-calculated for plug heave. The plug uplift contour line corresponds well with the recorded drop
in suction pressure at penetration of 20 m, if no contribution of reverse end bearing was taken into
account. In addition the plug heave was back-calculated by assuming a range of permeabilities to
assess the amount of heave. This showed again good agreement with the measured heave; however
it can be remarked that the variation in sand permeability leads to substantial differences in plug
heave predictions.

6.3 Numerical calculations

In order to verify the theory obtained within in previous analyses, numerical verification is done
using finite element software. Plaxis is a commonly used finite element-software tool for geotechnical
engineering, which provides a platform to assess complex conditions by solving them numerically.
Within the previous analyses, it turned out that reverse end bearing might be mobilized prior to plug
uplift. However, it is unknown to what extent and how long this additional ’capacity’ contributes,
which keeps the plug on its place during installation. From a practical point of view, it is interesting
to assess the plug stability, by giving a prediction of the mobilized reverse end bearing capacity. In
order to verify the theory in the most reliable way, the reference case of Chapter 2 has been taken,
with the specification as presented in Appendix F.3 and Table 6.5.
Table 6.5 Input-data for Plaxis calculations
Diameter Penetration zclay ksand kclay su γc0 Material Drainage

[m] depth [m] [m] [m/s] [m/s] [kPa] [kN/m3 ] Model type
10 10 3 1e−5 1e−6 10 8 MC Undr.(B)
According to the theory of Senders [2008], plug uplift for this case is likely to occur for pressure
differences higher than 36 kPa. This value is based on an adhesion factor (α) of 1 and excludes
possible contributions of reverse end bearing. This is also implemented in Plaxis by setting the
rigidity of the clay interface to 1. It is expected that for higher pressure differences, the effective
stresses below the clay plug become zero and uplift should occur. In the subsequent paragraphs a
further explanation of the uplift-mechanism in Plaxis is discussed.
Purpose of calculation
The purpose of the numerical verification is to gain insight in the time-dependent contribution of
reverse end bearing. Prior to draw conclusions regarding this topic, first the uplift-mechanism should
be identified with the use of Plaxis. This can be done by considering the vertical effective stresses just
below the clay plug. If the uplift mechanism is recognized with the use of Plaxis, the contribution of
reverse end bearing can be investigated.
6.3.1 Calculation method

Since Plaxis is based on continuum mechanics calculation procedures, it is not possible to model
a gap-filling problem in time, which is applicable during plug uplift. It is possible to model the
uplift-principle properly up to the moment that the plug will lift off. Beyond this point it is expected
that plastic points or dilatancy effects will intervene, because in that case a continuum medium is
-4.00 0.00 4.00 8.00 12.00 16.00 20.00 24.00 28.00 32.00 36.00 40.00 44.00
[m]
maintained. The input-model is presented in Figure 6.14, with horizontal dimensions of 40 m and
28.00
36
vertical dimensions of 20 m.
24.00 32
20.00 28
16.00 24
12.00
20
8.00
16
4.00
12
0.00
8
-4.00
Figure 6.14 Generated mesh of model 4
-8.00
0
The strategy of the modelling is to apply an increasing suction on top of the clay layer and monitor
the dissipation of pore pressures in time just below the clay plug. Therefore the stress-points (see
Figure 6.15) are chosen just below the clay plug as data-points. In addition data-points in the top of
the clay plug are taken to compare the pressure difference of the clay plug. Further details about the
calculation procedure are presented in Appendix F.3. The nodes A - F are defined to back-calculate
the deformation and pore pressures, nodes K,L and M are defined to back-calculate the (volumetric)
strains. For this reason a double set of data-points just below the clay plug is implemented.

Figure 6.15 Data-points to investigate behaviour of the plug

6.3.2 Modelling plug uplift

The chosen input model consist of an axisymmetric mesh which is refined close to the caisson skirt

and clay plug. Inside the caisson an artificial soil-material has been added, which has no effective
volumetric weight and therefore can be regarded as ’water’. This material is added to ensure a
continuum medium.

The principle of applying suction on top of the clay plug is complex using Plaxis. Therefore it

is chosen to apply a back-pressure on the surrounding soil, instead of applying suction inside the
caisson. This alternative has already be proven to model plug uplift, by the experiments that were
conducted in Chapter 3. It showed that for sufficient pressure difference the clay plug lifted off the
sand and uplift occurred due to the ’back-pressure’. For this reason the similar principle is adopted
for simulations using Plaxis.
Table 6.6 Calculation steps for Plaxis calculations
Step Water column in- Water column out- Differential Pressure incre- Time
side caisson [m] side caisson [m] pressure [kPa] ment [∆kPa] [min]
0 21 (initial) 21 0 0 0
1 20 21 10 -10 10
2 20 22 20 -10 20
3 20 25 50 -30 50
4 20 30 100 -50 100
5 20 40 200 -100 200
Within the Plaxis software, the differential pressure is obtained by setting the inner ’water’-
cluster of the caisson to dry cluster, while the surrounding soil is subjected to an phreatic head. In
the subsequent calculations, the phreatic head is increased to ensure an increasing pressure difference

over the clay plug. As a result, the differential pressure on the clay plug is gradually increased, which
is similar to installation-practice.
Table 6.6 presents an overview of the pressure increments with time. These values are based
on installation-data in practice according to Audibert et al. [2003]. The ’penetration-rate’ is kept
constant, by taking a suction increment for a fixed time period, as indicated in the right column of
Table 6.6. The penetration-rate is defined as suction/time, which is found to be around -1 kPa/min.
It should be noted that the increase of phreatic head is also applied to the underlying sand of the
surrounding soil, to avoid long-term consolidation over the top clay layer. In addition to this generic
example, three cases are considered to investigate the influence of installation-rate and cracking of
the clay plug. In order to investigate the time-dependency of installation, the time-period to equalize
the pressure is doubled, which makes the installation-rate be halved. In the subsequent analysis this
case is described by Case - 2t.
Output Version 2011.1.7671.7015 Output Version 2011.1.7671.7015
Project description Date Project description Date
Thesis_watertable 8/14/2013 Thesis_watertable 8/14/2013

(a) Cracking at the skirt (Case
Project filename
Thesis_watertable_crack
- crack )
User name
TU Delft
(b) Cracking in middle of clay (Case
Project filename
Thesis_watertable_crack_mid
- crack-mid )
User name
TU Delft
Figure 6.16 Cases to model cracking of clay plug
The remaining two cases on plug cracking are divided in cases with a sand-filled crack on different
locations. One case considers a crack along the skirt (Case - crack ), while the other case considers a
crack more in the centre of the clay plug (Case - crack-mid ). As can be noted in Figure 6.16, the crack
in Figure 6.16b is wider than the crack in Figure 6.16a. This is because the crack is modelled to be
an circular strip and therefore a smaller radius implies a wider gap to ensure the same fracture-ratio.
Both cases are based on a fracture-ratio of 10 %, which is in accordance to the theory of Chapter 4.
6.3.3 Results
This subsection discusses the results on several components, which are involved during plug uplift.
Subsequently the results of plug heave, reverse end bearing, installation-rate and cracking of the clay
plug are presented.
Plug heave
As stated in the introduction of this chapter, first the plug uplift mechanism should be identified.
This is done by considering the effective stresses just below the clay plug. Once the effective stresses
become zero just below the clay plug, it is considered that the resistances against uplift (self-weight
and skin friction) are overcome. The determination of the time when the effective stresses become
zero is shown for the initial case in Figure 6.17.
From Figure 6.17 it can be noted that the effective stresses just below the clay plug become zero
after 55 minutes, which corresponds with an applied suction of 55 kPa. In addition, the volumetric
strains are analyzed just below the clay plug. From Figure 6.17b it can be noted that from the moment
when the effective stress becomes zero, the volumetric strain of the underlying sand increases.

-25 0,20
0,18
K - initial
Vertical effective stress [kPa]

-20 0,16
K - initial L - initial
Volumetric strain [-]

0,14
-15 L - initial 0,12
0,10
-10 0,08
Phases Phase Phase
4 5 0,06
-5
1-3
0,04
0,02
0
0,00
0 50 100 150 200 0 50 100 150 200
Time [min] Time [min]
(a) Vertical effective stress (b) Volumetric strains
Figure 6.17 Output for data-points K and L according to Figure 6.15
From the findings of Senders [2008] it was found that the expected suction for plug uplift is
around 36 kPa. However, from the calculations with Plaxis it is found that at a suction of 55 kPa,
the uplift initiates. This is based on the fact that after 55 minutes the effective stresses become
zero, in combination with an increasing volumetric strain just below the plug at the same moment.
This is clearly an overestimation of the prediction according to Senders [2008]. However, this excludes
contributions of reverse end bearing. Further discussion about this is presented in the next paragraph.
Reverse end bearing
The second analysis is done on the generated capacity of reverse end bearing prior to uplift of the clay
plug. The magnitude of reverse end bearing is highly time-dependent due to dissipation of negative
excess pore pressures. The dissipation in time is interesting for installation-practice, since this will
describe the loss of an additional resistance against plug uplift. Figure 6.18 shows that for data-
points D and E, some negative excess pore pressures are found during installation. The vertical axis
presents the normalized excess pore pressure and the black dotted-lines indicate the applied pressure
difference over the clay plug.
1,5
Excess pore pressure / y'D [-]
D - initial (k = 6e-3 m/min)

E - initial (k = 6e-3 m/min)
1
Applied suction
0,5
0
0 50 100 150 200
-0,5
Time [min]
Figure 6.18 Excess pore pressures for data points D and E
As presented in section 5.1, the mechanism of reverse end bearing relies on suction mobilization
and suction release. This mechanism is also indicated in Figure 6.18 once an increase of the differential

pressure is applied at t = 0, 10, 20, 50 and 100 minutes. For data-points D and E it can be seen
that the excess pore pressures rapidly decrease due to the applied ’suction’ (the model assumes a
back-pressure). After this mobilizaiton of suction, it can be noted that the pressure increases to the
applied suction condition (suction release) which are presented by the black dotted-lines. For both
data-points it can be remarked that some negative excess pore pressures are generated due to the
pressure difference after 50 minutes and 100 minutes. After 55 minutes the generated mobilization
and release of suction has a duration of about 12 minutes, as can be noted in Figure 6.18. In this
time period the increment of suction equals -12 kPa, since the installation-rate is -1 kPa/min. This
implies that the applied suction is increased with 12 kPa, which results in a suction of 48 kPa. This
value for plug uplift is more or less found with the Plaxis results (55 kPa), which indicate that the
theory of Senders [2008] is an underestimation of the actual plug-stability.
2 2
D - initial (k = 6e-3 m/min) D - initial (k = 6e-3 m/min)
D - (k = 3e-3 m/min)
1,5 D - (k = 3e-3 m/min) 1,5
E - (k = 3e-3 m/min)

E - (k = 3e-3 m/min)
1 1
0,5 0,5
less permeable
0 0
0 50 100 150 200 95 105 115 125 135 145 155
-0,5 -0,5
Time [min] Time [min]
(a) Excess pore pressures for various permeabilties (b) Close up of suction mobilization and
for sand suction release
Figure 6.19 Excess pore pressures for data points D and E
By varying the permeability of the underlying sand, the contribution of the reverse end bearing in
time can be assessed (Figure 6.19). The permeability of the sand is lowered by a factor 2, which makes
the permeability equal to k = 5e−6 m/s. It shows that for a decrease of the permeability by a factor
of 2, the time period of suction mobilization and release is approximate twice the duration compared
to the higher permeable condition. Figure 6.19 also shows that for lowering the permeability, the time
period of reverse end bearing capacity increased. Figure 6.19b illustrates that for the case with lower
permeable sand the suction mobilization and release takes approximate 47 minutes, while for the
initial permeability it takes 20 minutes. The permeability of the underlying sand clearly influences
the dissipation of excess pore pressure during pressure differences acting over the clay plug. Therefore
the development of reverse end bearing in time can be related to the hydraulic characteristics of the
underlying sand.
Installation-rate
The theory can be verified by assuming an installation-time which is doubled, if no contribution of

reverse end bearing is considered. In theory this will also double the amount of plug heave, since this
is linear according to Chapters 4 and 5. The theory describes that the estimated plug heave will be
higher for lower installation-rates. This case assumes that the installation-time is doubled, which in
theory will also double the amount of plug heave.
From Figure 6.20 it can be seen that for the Case - 2t the heave is approximate 40 % higher for
a slower installation-rate. If the downward displacements of the caisson are also taken into account
it can be concluded that for the initial case an additional displacement of 7 cm was calculated. For
the for case - 2t the displacement of the caisson was 6 cm, which makes the relative displacements

0.20
0.18 D - initial
E - initial
0.16
D - 2t
0.14
Plug heave [m]

E - 2t
0.12
0.10
0.08
0.06
0.04
0.02
0.00
0 100 200 300 400
Time [min]
Figure 6.20 Calculated plug heave for lower installation-rate
of the plug 10 cm and 6 cm. This shows that the difference in relative displacement of the plug is
arount 50 % and thus in accordance to the theory.
Cracking
The influence of cracking of the clay plug is investigated by implementing two cases where a small
vertical gap of underlying sand material is considered (see Figure 6.16). From experimental research
is was found that a cracked plug will limit the plug heave due to the increased permeability of the clay
plug. As a result seepage flow can develop through the underlying sand and succesfull installation
can be achieved. However, according to the theory in section 4.1 it also shows that the amount of
cracking (fracture-ratio) is highly unlikely and uplift of the plug will intervene first.
0,14 D - initial
E - initial
0,12
D - crack skirt
Plug heave [m]
0,10 E - crack skirt

D - crack middle
0,08 E - crack middle
0,06
0,04
0,02
0,00
0 50 100 150 200
Time [min]
Figure 6.21 Calculated plug heave for cracked plugs
From Figure 6.21 it can be noted that calculated plug heave is lower for cases with a cracked plug
(10 % of cracks). Despite the small amount of uplift, it can be noted that uplift of data-points D and
E is lowered with 20 %. This reduction is lower than expected from the theory, but is likely being
caused by the limitations of the used software.

6.3.4 Discussion & conclusions

Plug stability is modelled by considering vertical effective stresses that become zero for sufficient
pressure difference acting over the plug. From data-points just below the clay plug, it turned out
that volumetric strains developed once uplift of the clay plug developed. On the one hand this is a
indication of plug uplift, on the other hand it is a limitation of the used software, since it based on
continuum mechanics and therefore does not model a water gap below the plug.
If this limitation is recognized, uplift of the clay plug can be identified by considering the vertical
effective soil stresses just below the clay plug. Once these become equal to zero, it can be assumed
that lift-off of the plug (i.e. plug instability) has occurred. In addition, dilatancy of the underlying
sand was encountered at the same moment, which also indicates plug uplift. From experiments in
Chapter 3, it was already found that the condition with back-pressure can be used to generate plug
uplift.
The principle of reverse end bearing relies on the dissipation of excess pore pressure in time, which
are identified for this Plaxis model. The suction mobilization and suction release are identified during
increasing pressure difference for a certain time period. As presented in the theory, the dissipation of
the reverse end bearing capacity is highly dependent on the permeability of underlying sand and the
pressure difference acting over the plug (applied suction). Quantification of the reverse end bearing
capacity is desired for assessment of the plug stability in time.
Both from theory and experiments it was found that a higher installation-rate will reduce the amount
of possible plug heave. The results of Plaxis show that a reduction of installation-rate by a factor
2 results in increase of residual heave by a factor 1,7. This dependency is nearly linear, which is in
accordance with the theory. It is therefore strongly recommended to consider this fact for installation
practice.
Cracking of the clay plug was investigated within the framework of this research by analytical cal-
culations and experimental tests. Both showed that a cracked plug will increase the permeability of
the plug, which will result in a pressure drop inside the caisson and increased seepage flow in the
underlying sand. The former contribution will limit the plug heave since the driving force to lift-off
the plug is lowered. The seepage flow will induce reduction of tip resistance, if the skirt is penetrated
into the underlying sand. Cracking of the plug will therefore reduce the residual plug heave. From
Plaxis results it is found that indeed the amount of plug heave is lower than for a solid plug, however
substantial reduction is not recognized. It is therefore recommended to perform more research to
the effects of cracking on the plug stability. Quantification can be achieved by laboratory tests with
suction cells or extension triaxial tests on layered soil samples.

6.4 Conclusions
For both field cases water injection devices were implemented as mitigating measure to decrease the
tip resistance in dense sands overlaid by less permeable material to a certain extent. This is not a
proven technology, and back-calculations did not clearly indicate the benefit of these devices. The
reduction of soil resistance and hence lowering of suction requirement could also be caused by the
seepage flow which was induced by uplift of the plug.
Plug uplift prediction
Plug uplift was observed physically in Case B, where the uplift suction requirement according to
the uplift theory of Senders [2008] was calculated. Considering the presented data of the paper
(Tjelta et al. [1986]) plug heave of 0.4 m was encountered during installation. Regarding the soil pro-
file, a potential plug for uplift failure was determined and back-calculated for plug heave. The plug
uplift contour line corresponds well with the recorded drop in suction pressure at penetration of 20 m.
From finite element calculations it was found that uplift of the plug resistance was underestim-
ated and lift-off initiated for higher suctions (or later when the reverse end bearing was dissipated).
The threshold of plug stability was modelled by considering vertical effective stresses that become
zero for sufficient pressure difference acting over the plug. From data-points just below the clay plug,
it turned out that volumetric strains developed once uplift of the clay plug developed. Consider-
ing the experiments in Chapter 3, it was already found that plug uplift can be generated by using
back-pressure instead of suction.
Cracking of the plug
For both cases a theoretical fracture-ratio can be back-calculated, where a circular strip as ’fracture-
gap’ is implemented. The surface of this gap is according to the theory of Chapter 4 and theoret-
ically ensures plug stabiliy by increasing the permeability of the clay plug. For both cases it can
be concluded that plug stability due to cracking is highly unlikely, since the required fractures were
unrealistic high (i.e. 10 - 20 %). However, possible contribution of reverse end bearing capacity was
not incorporated.
Cracking of the clay plug was investigated within the framework of this thesis by analytical cal-
culations and experimental tests. Both showed that a cracked plug will increase the permeability of
the plug and hence reduce the amount of plug heave. From the Plaxis results it is found that indeed
the amount of heave is lower than for a solid plug, however substantial reduction is not recognized.
It is therefore recommended to perform more research on the effects of cracking on the plug stability.
Quantification can be achieved by laboratory tests with suction cells or extension triaxial tests on
layered soil samples.
Estimation of plug heave
The amount of plug heave was back-calculated by taken into account a range of permeabilities. This
showed good agreement with the measured heave, however it can be remarked that the variation in
sand permeability leads to substantial differences in plug heave predictions. As well the installation
time is linearly dependent on the plug heave estimation.
The results of Plaxis show that a reduction of installation-rate by a factor 2 results in an increase of
residual heave by a factor of 1,7. Despite the difference in linearity, it still can be concluded that the
plug heave estimate is highly dependent on time-effects, i.e. permeability of the underlying sand and

installation time. It is therefore strongly recommended for installation practice to determine these
parameters properly and to assess the installation time carefully.
Contribution of reverse end bearing
The contribution of reverse end bearing as additional resistance against plug uplift relies on dissipa-
tion of negative excess pore pressures in time. Considering Case A it was found that the installation
time is relative short (i.e. approximate 1 hour) and the permeability of the underlying silty sand was
relatively low (1e−6 m/s). Therefore additional capacity of reverse end bearing was justified in the
back-calculations, which shows agreement with the drop in measured suction. The installation time
of Case B was relative slow (installation rate of 1 m/hr) and thus no capacity of reverse end bearing
was incorporated. Back-calculations showed that a good approximation of the uplift was obtained
by excluding the contribution of reverse end bearing.
The principle of reverse end bearing was also recognized for the Plaxis model, where the suction
mobilization and suction release were identified during increasing pressure difference for a certain
time period. As presented in the theory, the dissipation of the reverse end bearing capacity is highly
dependent on the permeability of the underlying sand. Predictions of the magnitude of reverse end
bearing are to be further researched.

Chapter 7
Conclusions & Recommendations
This Chapter presents the conclusions and recommendations on installation effects of suction caissons
in non-standard soil conditions, especially where sand is overlaid by clay. Section 7.1 presents the
conclusions of the research performed. In addition the recommendations for further research and
practical implementation are presented in Section 7.2.
7.1 Conclusions
Nowadays suction caissons are extensively used for anchoring large offshore installations at great
depths to the seafloor. Installation of suction caissons in coarse grained soils (with a high permeabil-
ity) is achieved by induced seepage flow in the soil. On the other hand for installation in fine grained
materials with a low permeability (silt or clay), no seepage flow will occur instead a differential pres-
sure used for installation will be mobilized. As a result the installation in layered soils, especially sand
overlaid by clay, is expected to be more challenging since the reduction of tip resistance is restrained
from flow restrictions in the clay layer. It is unclear to what extent the upper low permeability layer
will impact the seepage flow and thus the desired tip resistance reduction. In addition to this, the
plug stability should be assessed to avoid excessive plug heave.
Installation in layered soils
1. The favourable effect of water injection devices as mitigation measure to decrease the tip res-
istance of the suction caisson in dense sands is not fully supported by back-calculations of two
field cases. The apparent reduction of soil resistance and hence lowering of suction requirement
could also be caused other mechanisms (i.e. plug uplift, cracking of the plug or reverse end
bearing contributions) rather than only injection.
2. The design approaches for homogeneous sands (Houlsby and Byrne [2005] and Andersen et al.
[2008]) were successfully adjusted for calculation of the resistances in layered soils, where sand
is overlaid by clay. The most important adjustment was to correct the critical suction for the
homogeneous sandy material. The critical suction was horizontally translated for shallow penet-
ration in the sand, according to the solution of Senders [2008]. It should be noted, with regards
to the plug stability, that possible contribution of reverse end bearing is not incorporated.
Plug response during installation
1. Two different types of plug failure mechanisms are postulated for installation of suction caissons
in sand overlaid by clay; plug uplift and plug cracking. Considering several cases with D/z ≤ 6,
it was shown that plug uplift is more likely to occur during installation. However, there is an
uncertainty at the transition point regarding failure mechanism (i.e. cracking or uplift). From
81
schematization of a bending circular plate, an estimation of the cracking failure could be made
and showed that plug cracking is expected for D/z > 10, whereas uplift was expected for lower
D/z-ratios. This was supported by a practical case where uplift was observed at a lower D/z-
ratio (Case B). Comparing these two failure mechanisms it can be concluded that the failure
mechanism depends on the dimensions of the caisson and clay layer.
2. Experiments on plug cracking showed failure for high D/z-ratios, but also showed a pressure-
rate dependency. This corroborates data from the literature (Senders [2008] and Watson et al.
[2006]), where slow penetration rates highly contributes to uplift of the clay plug. Additionally,
high pressure-rates (fast installation) contributed to cracking of the clay layer during testing.
Therefore, the failure mechanism of the clay plug, i.e. plug uplift or plug cracking, depends on
the build up of suction pressure as a function of the desired installation-rate.
Assessment of plug stability
1. One of the scenarios to maintain a stationary plug during penetration of the suction caisson is
to increase the permeability in the upper clay layer (i.e. cracking). This results in adequate
tip resistance reduction in the sand and a reduced stress gradient across the clay-layer from
seepage flow such that plug uplift is prevented. The cracking-requirement for installation can
be calculated, by introducing a fracture-ratio. However, the resulting calculated values for
cracking are not realistic for typical installation practice, since at least 10 % of cracking area is
required for clayey soils. This conclusion was supported by two field cases (Chapter 6). It can
therefore be concluded that due to cracking only, it is highly unlikely that plug stability can be
maintained during installation. However, it should be remarked that no contribution of reverse
end bearing was considered.
2. When uplift can be accommodated, the relation between the plug velocity and the installation
rate should be assessed. Analytical analysis shows that for slow installation the total amount
of uplift is higher than for fast installation. The governing component that determines the
amount of uplift of the plug is the magnitude of seepage flow in the underlying sand layer (i.e.
permeability). Two case studies and numerical analysis show that the reduction of installation-
rate results in more plug heave. Furthermore, the plug heave estimate is strongly dependent
on time-effects, i.e. permeability of the underlying sand and installation time.
Contribution of reverse end bearing
1. For sands, the permissible capacity from reverse end bearing depends on the time required to
dissipate the negative excess pore pressures. This time requirement, defined as the time lag,
shows that for a typical installation period of 1 - 6 hours and hydraulic conductivities in the
sand layer lower than 1e−5 m/s some capacity due to delayed inflow of water can be considered.
It should be remarked that the chosen range of permeability has a great impact on the time
requirement for dissipation.
2. The underlying components for uplift of the plug were modelled using a spring-dashpot rheolo-
gical model. If the applied suction above the plug will initiate upward movement of the plug,
this movement will be restricted by the inflow of water below the plug. The difference in upward
movement between both components show similar response compared with mobilization and
release of reverse end bearing (i.e. disspation of negative pore pressures). The model indicates
that seepage flow increases rapidly for more permeable soils and therefore the time-period for
generating reverse end bearing capacity is significantly reduced. The mechanism of reverse
end bearing and the time dependency of this mechanism was identified in the installation data
of two field cases. A lower installation-rate seemed detrimental for the reverse end bearing

Conclusions & Recommendations 83
capacity. The principle of reverse end bearing was successfully modelled by the finite element
package PLAXIS. In the latter the suction mobilization and suction release are identified during
increasing pressure difference over a certain time period. As shown by the analytical model, the
dissipation of the reverse end bearing capacity is highly dependent on the hydraulic properties
of the underlying sand.

7.2 Recommendations
The theoretical analysis and the experimental work have shown the geotechnical mechanisms involved
for suction caisson installation in layered soils. This section presents recommendations for further
research and practical suggestions for suction caisson installations in sand overlaid by clay.
1. Preferred approach
Commonly two approaches are used to predict the soil resistance during penetration: (1) the
effective stress approach and (2) the CPT-approach. The first is an indirect method relying
on the friction angle and the effective unit weight of the soil, the second is a direct correlation
with the cone resistance. From an engineering point of view, the CPT-approach is preferred,
because the measured resistance is directly related to penetration resistance by a single factor.
This reduces the uncertainty in multiple input-parameters and avoids unnecessary elaborate
calculations. In addition, the stress-path of a CPT is comparable to the penetration of the
caisson skirt, which incorporates similar soil response in terms of (un-)drained behaviour.
2. Cracking of the plug
Cracking of the clay plug was investigated within the framework of this thesis by analytical
calculations and experimental tests. Cracks increase the permeability of the plug and hence
reduce the risk of plug heave. The analyses indicated that for less than 10 % cracks sufficient
seepage was achieved to limit plug uplift. Additionally, FE analysis showed that the plug heave
of a cracked plug is lower than for a solid plug. However, substantial reduction was not shown.
It is therefore recommended to perform additional research on the effect(s) of cracking on the
plug stability.
3. High installation-rate
When uplift can be accommodated to a certain extent, the relation between the plug velocity
and the installation rate showed that a higher installation-rate results in lower plug uplift
velocity and thus less plug heave in the same time period. The plug heave is highly dependent
on time-effects (permeability of the underlying sand and installation time). In order to minimize
plug heave, minimization of installation time (e.g. high pump capacity) and accurate (in-situ)
determination of the hydraulic properties is highly recommended.
4. Minimize the risk of installation-stop
When the caisson is halted during installation (e.g. due to breakdown of the pump), the under
pressure below the clay plug is maintained and thus seepage flow will continue. A sudden
installation-stop (once plug uplift has initiated) should therefore be prevented, since this will
cause ’maximal’ uplift of the plug. However, the plug will settle due to self-weight of the plug
once the pressure difference is decreased.
5. Contribution of reverse end bearing
Nowadays contribution of reverse end bearing is not often taken into account for layered soils.
The research presented shows that reverse end bearing capacity generally depends on the dissip-
ation of negative excess pore pressure in time. Since there is a time-effect involved, the capacity
of reverse end bearing greatly depends on the permeability of the underlying sand, drainage
length and installation time. For design purposes it is therefore recommended to maximize the
installation-rate and accurately determine the permeability of the underlying sand layer (e.g.
by dissipation tests). In cases of local shallow gas reservoirs, the dissipation of negative pore
pressures will develop quicker and thus the contribution in time will decrease.
6. Quantification of reverse end bearing
The theoretical analysis and numerical calculations within this thesis highlighted the possible
contribution of reverse end bearing in layered soil conditions. Further quantification can be
achieved by laboratory tests with suction cells or extension triaxial tests on layered soil samples.

Bibliography
Alhayari, S., K. Schroder, and A. Hill (1999). Curlew fpso: Successful suction anchor siting, design
and installation. Aberdeen.
Allersma, H., J. R. Hogervorst, and M. Pimoulle (2001). Centrifuge modelling of suction pile install-
ation in layered soil by percussion method. Brazil. ASME.
Andersen, K. H., H. P. Jostad, and R. Dyvik (2008). Penetration resistance of offshore skirted
foundations and anchors in dense sand. Volume 134, pp. 106–116. American Society of Civil
Engineers.
API (2000). Recommended Practice for Planning, Design and Constructing Fixed Offshore Platforms
- Working Stress Design. American Petroleum Institute.
Audibert, J. M., E. Clukey, and J. Huang (2003). Suction caisson installation at horn mountain–a
case history. In Proc Int Offshore and Polar Engineering Conf, pp. 762–769.
Bridge, C., K. Laver, E. Clukey, and T. Evans (2004). Steel catenary riser touchdown point vertical
interaction models. In Offshore Technology Conference.
Burns, C., P. Gauglitz, and R. Russell (2010). Shear strength correlations for kaolin/water slurries:
A comparison of recent measurements with historical data. Technical report.
Chen, J. and C. Juang (1996). Determination of drained friction angle of sands from cpt. Volume
122, pp. 374–381. American Society of Civil Engineers.
Cotter, O. (2009). The installation of suction caisson foundation for offshore renewable energy
structures. Technical report.
DnV (1992). Offshore standard dnv-os-j101 - design of offshore wind turbine structures.
Erbrich, C. and T. Tjelta (1999). Installation of bucket foundations and suction caissons in sand. In
Offshore technology conference.
Feld, T. (2001). Suction buckets, a new innovative foundation concept, applied to offshore wind
turbines.
Gourvenec, S., H. Acosta-Martinez, and M. Randolph (2009). Experimental study of uplift res-
istance of shallow skirted foundations in clay under transient and sustained concentric loading.
Géotechnique 59 (6), 525–537.
Houlsby, G. T. and B. W. Byrne (2005). Design procedures for installation of suction caissons in clay
and other materials. Proceedings of the ICE-Geotechnical Engineering 158 (2), 75–82.
Huang, J., J. Cao, and J. M. Audibert (2003). Geotechnical design of suction caisson in clay. In
Proceedings of ISOPE-2003: Thirteenth (2003) International Offshore and Polar Engineering Con-
ference; Volume 2, pp. 770–779.
85
Hvorslev, M. J. (1951). Time lag and soil permeability in ground-water observations.
Randolph, M. and M. R. S. Gourvenec (2011). Offshore geotechnical engineering. Taylor & Francis.
Roderick, G. and A. Lubbad (1975). Effect of object in-situ time on bottom breakout. In Offshore
Technology Conference.
Senders, M. (2008). Suction caissons in sand as tripod foundations for offshore wind turbines. Tech-
nical report.
Senders, M. and M. F. Randolph (2009). Cpt-based method for the installation of suction caissons
in sand. Volume 135, pp. 14–25. American Society of Civil Engineers.
Senpere, D. and G. Auvergne (1982). Suction anchor piles-a proven alternative to driving or drilling.
In Offshore Technology Conference.
Thusyanthan, N., W. Take, S. Madabhushi, and M. Bolton (2007). Crack initiation in clay observed
in beam bending.
Tjelta, T., T. Guttormsen, and J. Hermstad (1986). Large-scale penetration test at a deepwater site.
In Offshore Technology Conference.
Tjelta, T. I. (2001). Suction piles: their position and application today. In Proc., 11th Int. Offshore
and Polar Engineering Conf., International Society of Offshore and Polar Engineers, Mountain
View, CA.
Tran, M. (2005). Installation of suction caissons in dense sand and the influence of silt and cemented
layers. Technical report, Sydney.
Tran, M., M. Randolph, and D. Airey (2007). Installation of suction caissons in sand with silt layers.
ASCE.
Vermeer, P. and C. Meier (1998). Stability and deformations in deep excavations in cohesive soils.
In Proceedings International Conference on Soil-Structure Interaction in Urban Civil Engineering,
Darmstadt Geotechnics, Volume 1.
Verruijt, A. and S. van Baars (2007). Soil mechanics. VSSD.
Watson, P., C. Gaudin, M. Senders, and M. Randolph (2006). Installation of suction caissons in
layered soil.

Appendices
87
Appendix A
Implementation of effective stress

and CPT-approaches
Suction caissons in sand as tripod The University of Western Australia
foundations for offshore wind turbines School of Civil and Resource Engineering
The first term was unfortunately wrongly copied into Houlsby and Byrne (2005) and
A.1missed
Calculations according to effective stress-approach
the term k in the bottom of the equation. However, the calculations in their
fac
paper
For the were performed
effective with the correct
stress-approach (Houlsbyexpression.
and Byrne [2005]) the method assumes a linear reduction
The authors assumed that the influence of the
in internal friction and end bearing and an suction,
increase in p,outside
on the friction
internal and
when external
suction is applied. A
pore pressure factor is
stress conditions (a)linear
is calculated
over depth.which accounts
The stress for the
conditions 3D-effect
during suctionofcan
thetherefore
suction caisson.
be expressed by:
0 dσ vo ' σ 0 σ vo ' p
γ '+− 1−−(1a(z)
− a( z ) )S

dσvo
= γs0 + = vo (A.1)
dz dz Zo Z o zz
dσ vi ' σ ' p (2.7)
= γ '+ vi + a( z )
dz Zi z
0 0
dσvi σ S
= γs0 + vi + a(z) (A.2)
The determination of the zstress level at tip level is again not straightforward. The
dz Z i
Ininternal
static and external the
conditions stress will change
internal with issuction
pressure pressure
typically and but
larger, whilewith
the internal
increasing pressure the
pressure is typically larger in static conditions, the external stress level will become
external stress level will become larger. The stress level at the tip is therefore calculated by taking
larger where
the difference into(sufficient)
account; suction is applied. The stress at tip level must therefore be
calculated depending on the difference of internal and external stress:
⎧
⎪ σ vi ' N q + γ ' wt N γ in case (σ vo '−σ vi ') ≥ 2wt N γ / N q
⎪⎪
σ end ' = ⎨ σ vo ' N q + γ ' wt N γ in case (σ vi '−σ vo ') ≥ 2wt N γ / N q
⎪ ⎛ 2x 2 ⎞
⎪σ vo ' N q + γ ' ⎜⎜ wt − ⎟⎟ N γ in other cases
⎩⎪ ⎝ wt ⎠
with (2.8)
w (σ vo '−σ vi ')N q
x= t +
2 4γ ' N γ
The reduced stress

2.4.3 CPTis presented
approachesin the equation below; as well the reduced force can be determined.
Three CPT methods are presented in this section: the DNV method, Feld’s method and
the NGI method. The(z)
Wtot,Houlsby,F DNV
= αmethod only describes the self weight penetration phase
LOW out (z)Qoutside + αin (z)Qinside + αtip (z)Qtip (A.3)
whereas the other two methods describe both the self weight penetration phase and the
suction penetration phase.
DNV method
αout (z) = [1 + a(z)]Sred Ai (A.4)
The current DNV method uses two dimensionless parameters, kf_c and kp, to relate the
unit friction and unit end bearing respectively to the89 cone penetration resistance, qc. The
caisson installation resistance, Rc, can be expressed as the sum of inner friction, Fi, outer
friction, Fo, and tip resistance, Qtip, according to (DNV, 1992):
αin (z) = [1 − a(z)]Sred Ai (A.5)
αtip (z) = f (αout , αin ) (A.6)
A.2 Calculations according to CPT-approach

Suction caissons in sand as tripod The University of Western Australia
foundations for offshore wind turbines School of Civil and Resource Engineering
The CPT-approach of DnV is not used to predict the changes in soil resistance during suction assisted
penetration. However, Andersen et al. [2008] presented a calculation model based on critical pressure,
suctionSpressure,
number, self-weight, penetration depth and wall thickness. Senders [2008] developed a best-fit
N,cr, upon the permeability ratio, kfac_thin, and the relative penetration, z/D
equation to describe the dependence of the critical suction number (SN,cr = pcrit /zγ 0 ).
according to:
⎛ z ⎞
⎛ ⎛ z ⎞ 0.5 ⎞ k fac _ thin 0.675−0.3⎜⎝ D ⎟⎠ z
S N ,cr ≈ 0.16⎜ 5⎜ ⎟ + 3 ⎟ for >0
⎜ ⎝D⎠ ⎟ ⎛z⎞
0.5
D (2.12)
⎝ ⎠ ⎜ ⎟
⎝D⎠
FigureThe
2-5bpermeability-ratio (kf ac thinfits=the
shows that this equation ki /k o ) ispresented
data calculated
byby taking aetthin
Andersen slice directly next to
al. (2008)
the caisson wall, which is assumed to have a different permeability during suction. With the ratio
well.
Rc,F low /Rc the following equation can be used to calculate the reduction in overall soil resistance;
To determine the ratio between the resistance during ‘no flow’ and ‘flow’ conditions a
graphical method has to be used. Empirically it!was 1 found that the ratio between the
Sreq a(z)
a(z)
resistanceQtot,N
during suction
GI,F LOW (z) =installation,
1− Rc,flow, and [Q the resistance
outside during
(z) + Qinside push
(z) + in
Qtip (z)] (A.7)
Scrit,N GI
installation, Rc, depends on the ratio between the penetration and the wall thickness,
z/wt, and the ratio between the suction number and the critical suction number, SN/SN,cr
A.3 Correlation of effective stress- and CPT-approaches
(Figure 2-6a). Again a best fit equation was developed for the data presented in this
graph
Fortothe
facilitate
model calculations
the followinginparameters
the remainder
whereof this
used;thesis:
kf = 0.0015, kp = 0.45 (most probable) and
kf = 0.003, kp = 0.6 (highest 1 expected) qc =20 MPa, su =50 kPa, α = 0.4. From Figures A.1 and
A.1 it Rcan be ⎛
seen S
that ⎞
a a
the correlations of the Houlsby-method are close to the highest expected
≈ ⎜⎜1 − N ⎟⎟
c , Flow
estimates.
R S
c ⎝ N ,cr ⎠
where
(2.13)
1.5
z
a = 0.01
wt
Figure 2-6b shows that this equation fits the presented data from Andersen et al. (2008)
reasonably.
For each penetration depth, z, the critical suction number, SN,cr, needs to be determined
and it needs to be checked whether the self weight is higher than the push-in resistance,
Rc. If this is not the case, Figure 2-6 is used to determine the resistance and the suction
pressure. This is done by entering the graph at the ratio of the self weight and the push-
in resistance, W'/Rc (in Figure 2-6a arbitrarily chosen to be 0.3). From this point
onwards the grey dotted line, which has a gradient of 1:0.25πD2zγ'SN,cr/Rc is followed
until the penetration ratio z/wt is encountered. From the intersection of the line and the
curve it can be determined what the Rc,flow/Rc-ratio and the SN/SN,cr-ratio are. This can be
used to calculate
R.H. Romp the penetration resistance and necessary suction pressure. M.Sc. Thesis
To summarise this NGI method in formula form, the following equations are used:
Appendix 10,00 91
Inner resistances (kN) Outer resistances (kN)

0 1000 2000 3000 4000 0 1000 2000 3000 4000
0,00 0,00
Houlsby & Byrne Houlsby & Byrne
1,00 NGI (most probable) 1,00

NGI (most probable)
NGI (highest expected) NGI (highest expected)
2,00 2,00
3,00 3,00
4,00 4,00
Depth (m)
Depth (m)
5,00 5,00
6,00 6,00
7,00 7,00
8,00 8,00
9,00 9,00
10,00 10,00
(a) Inner resistances (b) Outer resistances
Tip resistances (kN) Total soil resistances (kN)

0 5000 10000 15000 0 5000 10000 15000 20000
0,00 0,00
Houlsby & Byrne Houlsby & Byrne

1,00 1,00
NGI (most probable) NGI (most probable)
NGI (highest expected) NGI (highest expected)

2,00 2,00
3,00 3,00
4,00 4,00
Depth(m)
Depth (m)
5,00 5,00
6,00 6,00
7,00 7,00
8,00 8,00
9,00 9,00
10,00 10,00
(c) Tip resistances (d) Total resistances
Figure A.1 Estimated inner-, outer, tip- and total resistances according to Houlsby and Byrne [2005] and
Andersen et al. [2008]


The earth pressure in the deep part of the anchor, where the increase with depth
soil is assumed to flow around the anchor, is based on shear strength at a
solutions from the method of characteristics of a perfectly more conservative
smooth and a fully rough wall /14/. However, the solutions are
corrected for the effect of the varying adhesion factor at the In case of an open
skirt wall/soil interface αout and the vertical shear stress at the instead of the vert
skirt wall. The correction factors are found by finite element if the inside skirt f
analyses. The depth to the deep part of the anchor H1 is equal
to the depth where the resultant horizontal stress from active Due to coupling b
and passive earth pressures, including shear stress at the side of
the anchor, becomes larger than the earth pressure from the the skirt wall and t
failure mechanism corresponding to soil that flows around the shear stress at the
anchor. bearing capacity, t
Appendix B numerical optimis
--- End of Guidance Note ---
of such a procedur
The inverse bearing capacity Vanchor,tip at the bottom of the anchor resistance c
the equilibrium co
Determination of end-bearing
anchor model in Figure 4-2 is based on Brinch-Hansen’s
bearing capacity equations /15/. For pure vertical loading eye Tp at failure (a
the inverse bearing capacity below skirt tip can be the anchor W’ is in
coefficient calculated with a bearing capacity factor ranging from the pad-eye zp is fo
Nc=6.2 at the surface to Nc=9 at depths greater than 4.5 moment acting on
times the diameter, see Figure 4-4, which is slightly more zero:
Rh = T p ⋅ cos(α p )
conservative than recommended in /15/. The expression
For pure vertical loading
for thetheNinverse bearing capacity below skirt tip can be calculated with a bearing
c-factor in Figure 4-4 is
capacity factor ranging from Nc = 6.2 at the surface to Nc = 9 at depths greater than 4.5 times the
⎛
diameter, see Figure B.1. The expression i ⎞⎞
for the N⎛cz-factor
N c = 6 .2 ⋅ ⎜⎜ 1 + 0 .34 ⋅ arctan ⎜ ⎟ ⎟⎟
in Figure B.1 is; Rv = Tp ⋅ sin α p ( )
⎝ ⎝ D ⎠⎠ (4-1)
z ! M soil = T p ⋅ cos(α
valid for zi i≤ 4 .5 valid f or zi = 4.5
Nc = 6.2 1 + 0.34arctan (B.1)
DD D where
which
which includes a shape includes
factor a shape factor sc
sc = 1.2. = 1.2. Tp = line
the
NC the
0,0
6,0 7,0 8,0 9,0
αp = loa
loa
xp = hor
of t
Normalised depth, zi /D
1,5
zp = dep
H = ski
3,0 Rh = hor
pad
Rv = ver
eye
4,5
res
sub
Figure B.1 Msoil
Bearing capacity factor Nc vs normalised depth zi /D for circular foundation with pure vertical = res
loading Figure
(Fig 4-4 4-4
from Bearing capacity factor Nc vs normalised depth
DnV [1992]) abo
zi/D for circular foundation with pure vertical loading.
The main results f

Figure 4-2 are the
can be taken by th
padeye depth zp.
DET NORSKE VERITAS
93

Appendix C
Prediction method to determine

Literature Review and Experiments Remon Romp
cracking of the plug
Appendix 1: Plug cracking calculation
Calculation of plug cracking due to bending moment;

A calculation of plug cracking due to bending moment:
p crack
su
z
Maximum circular bending moment

Figure for pressureofper
C.1 Schematization plugunit width:
cracking
Maximum circular bending moment for pressure per unit width:
(1 + ν)pcrack Di2
Tmax,clamped = [kN m/m] (C.1)
Source: http://www.me.ust.hk/~meqpsun/Notes/Chapter3.pdf
64
Soil resistance per unit width:

(3 + ν)pcrack Di2
Tmax,hinged = [kN m/m] (C.2)
64
Source: Thusyanthan (2007) Crack
Source: http://www.me.ust.hk/ Initiation in clay observed in beam bending
~meqpsun/Notes/Chapter3.pdf
(A = zclay*B) and B = 1m.
Soil resistance per unit width;
Range of solutions; (i.e. first clamped then hinged)
2
Tmax,clay = su zclay A = su zclay [kN m] (C.3)
Equation C.3 is according to Thusyanthan et al. [2007], with A = zclay B and B = 1 m.
Range of solutions; (i.e. first clamped then hinged)
2 2
64su zclay 64su zclay
< pcrack < (C.4)
πDi3 (3 + ν) πDi3 (1 + ν)
95

Appendix D
Installation of a suction caisson
D.1 Steady-state suction

Depending on the permeability of the clay plug, steady-state suction can be reached in that case.
The pressure increment due to seepage flow equals the pressure decrease due to suction as can be
seen in Figure D.1a.
-20
δSout
δt -18
-16
-14
Suction pressure [kPa]
δSin -12
δt -10
Suction without seepage
-8 Suction with seepage | k=1e-3
Suction with seepage | k=1e-4
hc -6 Suction with seepage | k=1e-5
-4 Suction with seepage | k=1e-6
Critical suction
-2
Sand 0
0 1 2 3 4 5 6
Time[hours]
(a) Schematization of steady-state suction (b) Reduction of underpressure due to seepage flow (z = 2 m)
Figure D.1 Illustration of steady-state suction during installation
Figure D.1b shows that steady-state suction is reached within 1 hour for soil with permeability of
1e−3 m/s (no volumetric strain or dilatancy effects). It can also be remarked that for permeabilities
lower than 10 e−5 m/s (clays) steady state suction will not be reached for common installation
periods.
97
D.2 Elastic plug heave

The proportion of the caisson wall that is accommodated by the flow of soil inwards into the caisson
or outwards is discussed Randolph and Gourvenec [2011]. He presented that a 50:50 split during self-
weight penetration (Figure D.2 - left), but a 100:0 split during suction assisted penetration (Figure
D.2 - right).
Figure D.2 Soil flow around caisson tip (fig 7.35 of Randolph and Gourvenec [2011])
Despite the stage of installation (self-weight or suction assisted), it can be remarked that there
will be a contribution of internal plug heave due to the installed caisson. In case of sand overlaid by
clay, there will be limited reduction at the tip and therefore limited inward flow of the sand. A 50:50
split is taken for layered soil, both for self-weight penetration as for suction assisted penetration. By
assuming elastic behavior of the soil, the amount of plug heave depends on the penetrated volume of
caisson.
hheave πD2
∆Vcaisson = ∆Vheave = 0.5Atip ∆z = (D.1)
4
2Atip ∆z
hheave = (D.2)
πD2
From equation D.2 it can be seen that the amount of plug heave linearly relates to the penetration
depth (z ). If the tip-surface is estimated as the circumference times the thickness (t), the heave can
be estimated by;
hheave 2t
≈ (D.3)
z D

Appendix 99
D.3 Procedure to determine fracture-ratio

1. Determine soil-resistances
Modified Houlsby & Byrne and DnV-methods are implemented, according to respectively ef-
fective stress- and CPT-approaches.
2. Correlate effective stress- and CPT-approaches

For the model the following parameters where used; kf = 0.0015, kp = 0.45 (most probable)
and kf = 0.003, kp = 0.6 (highest expected) qc = 20 MPa, su = 50 kPa, α = 0.4.
3. Calculate Scrit
According to the methods presented, a critical suction pressure should be determined. This
critical suction will be the upper bound of the applied suction. The difference between the
approaches to calculate the critical suctions can be found in Figure 2.9.
4. Calculate Sred
According to the methods discussed in previous section, the required suction for installation
can be calculated. This suction is sufficient to induce a seepage flow in the underlying sand, and
thus should be applied just below the clay plug. The true suction in the can (S ) is calculated
in the next step.
5. Back-calculate kcrit
The true suction S (which is induced by the pump) is limited by the critical suction in case of
plug uplift failure. This upper boundary for the pressure-ratio Scrit /Sred is then back-calculated
to the critical permeability (kcrit ) according volume continuity, which implies that the specific
flow through the sand have to be equal of that through the clay plug. The permeability for the
underlying sand is assumed to be ksand = 1.0 e−4 m/s.
6. Determine Ao /Abase
The critical permeability can be expressed in terms of a fracture-ratio. This presents the ratio
of cracks in the clay, which is required for installation in layered soil.


Appendix E
Modelling on plug stability
E.1 Determination of time lag

Rate of inflow at time t, based on theory of Hvorslev [1951]
S
Qin = k Ai (E.1)
γw s
Volume of inflow in time dt;
Qin dt = Ain dz (E.2)
And;
γw dz Ai
= k dt (E.3)
S s
The total flow for equalization of pressure difference is defined by;
S
V = Ain (E.4)
γw
And;
Ain S s
T = = (E.5)
Qin γw k
101
Installation of Suction Caissons in Layered Soils Remon Romp
Appendix 7: Schematization of plug uplift
E.2 Schematization of plug uplift
Pressures
Pressures and volumes
and volumes in time:
in time;
TIME 
Differential
Suction S1 applied
pressure on plug
during installation
increases
S1 = Splug,uplift T=0
+ ΔS T=1
S1 > Splug,uplift S2> 0
Attraction of water
Fres = m*a due to S2
Δs = ½ *a*Δt2 Inflow of water due

T=2
to S2
V1 decreases V2 decreases
T=4
Reduction of S1 Reduction of S2
S1<Splug,uplift S1>Splug,uplift S2 = 0 S2>0 T=5
Delft University of Technology 71 SBM Offshore

Appendix F
Parameters for verification
F.1 Case A: Curlew FPSO
(a) Anchor 1 (b) Anchor 2
(c) Anchor 3
Figure F.1 Parameters for calculations of Curlew case (Case A)
103
F.2 Case B: Gullfaks C
Figure F.2 Soil data of Gullfaks C (Case B)
Figure F.3 Clay content of Gullfaks C (Case B)

Appendix 105
F.3 Plaxis calculations

Material model
For both soil types (clay and sand) the Mohr-Coulomb model is taken as first approximation.
According to the Plaxis 2D manual it can be found that the Mohr Coulomb model is based on
a first order approximation of soil behaviour and includes a limited number of features. Care
must be taken in undrained conditions, since the effective stress path may not be realistic.
Alternatively, the model may be used with friction angle φ set to 0 and the cohesion set to su ,
to enable a direct control of undrained shear strength.
Drainage type
Drained analysis are appropriate for conditions when the permeability is high, the rate of loading
is low and short term behaviour is not of interest. Undrained analysis are appropriate for low
permeability soils, fast rate of loading and long term behaviour conditions. A suggestion by
Vermeer and Meier [1998] for deep excavations presents undrained analysis for T < 0.10 and
drained analysis for T > 0.40 according to equation F.1.
kEoed
T = t with; (F.1)
γw D 2
k = Permeability [m/s]
Eoed = Oedometric modulus [kPa]
γw = Unit weight of water [kN/m3 ]
D = Drainage length [m]
t = Installation time [s]
T = Dimensionless time factor [-]
The parameter D represents the drainage length and is calculated according to the solution
of Senders [2008], which is presented in equation 2.15. Within the framework of this thesis it
can be found that drained analysis are applicable for the underlying sand, whereas undrained
analysis for the clay material is appropriate. Since the sand material is considered to behave
drained, build-up of excess pore pressures are not considered. However, the negative excess
pore pressures are of interest, since reverse end bearing is considered. Using the Mohr-Coulomb
model with the undrained (A) method overestimates the undrained shear shrength of soft clays.
In order to use su as an input parameter it is preferred to use the undrained (B) method.
Calculation procedure
Classical mode is based on Terzaghi, advanced mode base on Bisshop, which takes into ac-
count suction for unsaturated soils. The latter is not applicable since fully saturated soils are
considered, therefore classical mode has been implemented for calculations.
Furthermore the Consolidation (EPP)-calculation procedure is adopted to analyse the devel-
opment of dissipation of excess pore pressures in a saturated clay-type soil as function of time.
Pore pressures
Pore pressures in stress points and external water pressures at model boundaries are updated
during the calculation according to the deformed model boundaries and the displaces position
of stress points. Basis for the update of water pressures is the general phreatic level and the
cluster phreatic levels. In this way, the buoyancy effect of soil that is submerged below the
phreatic level is taken into account.

Material properties
Figure F.4 Material properties of Sand (general)

Appendix 107
Figure F.5 Material properties of Sand (parameters)

Figure F.6 Material properties of Sand (flow parameters)

Appendix 109
Figure F.7 Material properties of Clay (general)

Figure F.8 Material properties of Clay (parameters)

Appendix 111
Figure F.9 Material properties of Clay (flow parameters)

Output Version 2011.1.7671.7015
Printscreens of groundwater head due to applied suction
(a) Step 1
Project description Date
Thesis_watertable 8/21/2013
Project filename Step User name
Thesis_watertable 4 TU Delft
(b) Step 2
Figure F.10 Calculation steps 1 and 2 for initial case


Appendix 113
(a) Step 3
(b) Step 4
(c) Step 5
Figure F.11 Calculation steps 3 - 5 for initial case



Remon Romp - Thesis (Final) HQ

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Remon Romp - Thesis (Final) HQ

Uploaded by

Copyright:

Available Formats

Delft University of Technology

Installation-effects of suction caissons

Delft University of Technology

8th October 2013

Dr. Ir. J. Dijkstra - Daily supervisor University

Ir. S. Frankenmolen - Daily supervisor Company

Dr. Ir. K.J. Bakker - External member

Delft University of Technology SBM Offshore

Stevinweg 1 24 Avenue de Fontvieille

8th October 2013,

Remon Hendrik Romp

Administrative Data iii

2 Installation in layered soils 5

3 Plug response during installation 21

3.2 Experiments on plug response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4 Assessment of plug stability 37

5 Modelling components of plug uplift 47

6 Verification of the theory 59

7 Conclusions & Recommendations 81

A Implementation of effective stress and CPT-approaches 89

B Determination of end-bearing coefficient 93

C Prediction method to determine cracking of the plug 95

D Installation of a suction caisson 97

E Modelling on plug stability 101

F Parameters for verification 103

2.1 Self-weight penetration and suction-assisted penetration during installation . . . . . . 5

3.1 Schematization of plug uplift parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4.1 stable and moving plug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.9 Fracture ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

5.1 Components of plug uplift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

C.1 Schematization of plug cracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

D.1 Illustration of steady-state suction during installation . . . . . . . . . . . . . . . . . . 97

F.1 Parameters for calculations of Curlew case (Case A) . . . . . . . . . . . . . . . . . . . 103

F.3 Clay content of Gullfaks C (Case B) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

3.1 Scaling laws for 1G testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

6.1 Key-data for Case A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

Symbol Significance Unit

a = Parameter after method of Andersen et al. [2008] [-]

Symbol Explanation Unit

Qclay = Flow through clay [m3 /s]

1.1 Background information

1.2 Problem description

Water out Water out

(a) Installation in sand (b) Installation in clay

Figure 1.2 Installation in homogeneous soils

Water out Water out

Figure 1.3 Installation in layered soils

R.H. Romp M.Sc. Thesis

1.3 Purpose of research

- Comparison of the analytical model with a commercially available geotechnical finite-element

- Only installation in layered soils is considered;

- A simplified geometry for the suction caisson is considered;

M.Sc. Thesis R.H. Romp

1.4 Readers manual

R.H. Romp M.Sc. Thesis

Installation in layered soils

2.1 Installation of a suction caisson

Lower caisson + suction pump Caisson touchdown & initial

Pump water out caisson untill target Undock of suction pump

Figure 2.1 Self-weight penetration and suction-assisted penetration during installation

Water out Water out

(a) Installation in permeable soil (b) Installation in low permeable soil

2.1.1 Reduction of tip resistance in layered soils